core-admin command reference
The operator CLI, shipped in core-runtime. Generated from the command tree in this repository; see the CLI overview for which binary does what.
108 commands.
| Group | Commands | Purpose |
|---|---|---|
admin |
12 | System forensics: decision traces, refusals, and pattern analytics. |
capabilities |
1 | Inspect and search CORE capabilities. |
code |
5 | Codebase quality, style, and verification operations. |
cognitive-roles |
1 | Cognitive-role capability projection (#821 Unit 2). |
coherence |
8 | Constitutional Coherence Checker — scan ADRs, rule domains, and northstar documents for candidate contradictions, gaps, and drift. Governing ADR: ADR-067. |
commands |
1 | Search and inspect registered CLI commands. |
constitution |
4 | Operations for the system mind: policies, schemas, and rule coverage. |
context |
3 | Build and explore context packages for LLM assistance. |
daemon |
6 | Background worker daemon management. |
database |
6 | PostgreSQL state and operational data management. |
decisions |
2 | Decision history: the repository's ADRs, with their files (ADR-168). |
demo |
2 | Isolated, opt-in demonstrations that CORE governs its own changes. |
dev |
11 | High-level developer workflows: synchronization, refactor, and stability tools. |
docs |
1 | Reader-facing documentation generated from live sources (the CLI trees). |
grc |
2 | Compliance gap-analysis: check a document corpus against a requirements catalog. |
interactive-test |
2 | Interactive test generation with step-by-step approval |
law |
9 | Grounded answers about this repository's law, with sources (ADR-168). |
llm-resources |
2 | core.llm_resources authoring/validation surface (#821 Unit 3). |
mcp |
1 | The assistant surface over MCP: law, decisions and verdict for any AI assistant (ADR-168). |
project |
3 | Operations for project lifecycle: scaffolding. |
refactor |
4 | Refactoring analysis and suggestions |
runtime |
4 | Runtime state and health of the running CORE system. |
secrets |
4 | Manage encrypted secrets in the database. |
status |
1 | Single-glance system state and readiness. |
symbols |
1 | Operations for the symbol registry and Knowledge Graph identification. |
tests |
1 | Test maturity dashboard and autonomous generation insights. |
tools |
2 | Governed, operator-focused maintenance and refactoring tools. |
vectors |
3 | Vector store operations (Qdrant). |
workers |
6 | Constitutional worker management. |
core-admin admin
System forensics: decision traces, refusals, and pattern analytics.
core-admin admin accessibility
core-admin admin accessibility [OPTIONS]
Derived census of CLI command and API route exposure classifications. Read from @command_meta.exposure and ROUTER_EXPOSURE module constants. Filter with --tier user-facing or --tier governor-only.
| Argument / option | Default | Description |
|---|---|---|
--tier |
Filter by exposure tier: user-facing or governor-only | |
--format |
table |
Output format: table or json |
core-admin admin coverage
core-admin admin coverage
Generate a Traceability Matrix. Identifies 'Critical Gaps' where Blocking rules have no enforcement engine.
core-admin admin forensics
core-admin admin forensics SESSION_ID
Reconstructs the history of an autonomous operation. Shows exactly what the AI thought vs what the System physically did.
| Argument / option | Default | Description |
|---|---|---|
SESSION_ID |
required | The UUID of the session to investigate. |
core-admin admin health
core-admin admin health [OPTIONS]
Reads the system_health_log populated by the continuous ObserverWorker. Replaces the legacy 'check audit' active scan with a passive, instant ledger read.
| Argument / option | Default | Description |
|---|---|---|
--limit, -n |
10 |
Number of recent health snapshots to show. |
core-admin admin legacy
core-admin admin legacy [OPTIONS]
Find every HEALED, CONSTITUTIONAL FIX, CIRCULARITY FIX, DEPRECATED, FUTURE, PENDING, HACK, WORKAROUND and LEGACY marker in the codebase.
This is your cleanup map. Start with high severity, highest count files.
Examples: core-admin admin legacy core-admin admin legacy --severity high core-admin admin legacy --marker HEALED --detail core-admin admin legacy --top 5 --detail
| Argument / option | Default | Description |
|---|---|---|
--severity, -s |
Filter by severity: high, medium, low | |
--top, -n |
20 |
Show top N files by debt load (default: 20) |
--detail, -d |
off | Show individual lines, not just file summary |
--marker, -m |
Filter by marker type (e.g. HEALED, FUTURE, DEPRECATED) |
core-admin admin patterns
core-admin admin patterns [OPTIONS]
Analyze architectural pattern usage and classification accuracy.
Helps identify if certain code patterns (like 'action_pattern') are frequently misidentified or failing validation.
| Argument / option | Default | Description |
|---|---|---|
--last, -l |
10 |
Number of recent traces to analyze. |
core-admin admin refusals
core-admin admin refusals [OPTIONS]
Audit constitutional refusal outcomes.
Refusals are first-class outcomes where CORE chose to stop because an operation was unsafe or lacked sufficient context.
| Argument / option | Default | Description |
|---|---|---|
--limit, -n |
20 |
|
--type, -t |
Filter by type (e.g. boundary, extraction) |
core-admin admin self-check
core-admin admin self-check [OPTIONS]
Validate CLI registry against structural and constitutional rules. Use --write to automatically assign missing @command_meta decorators.
| Argument / option | Default | Description |
|---|---|---|
--verbose, -v |
off | Show full command listing. |
--write |
off | Autonomously repair missing metadata using AI. |
core-admin admin status
core-admin admin status
Perform a sensory check of the Body's physical connections. Validates Database, Vector store, and Environment coordinates.
core-admin admin summary
core-admin admin summary [OPTIONS]
Summarizes the most recent actions in the Body's ledger. Explicitly highlights 'Pain Signals' (errors) for human review.
| Argument / option | Default | Description |
|---|---|---|
--limit, -n |
15 |
Number of recent actions to analyze. |
core-admin admin traces
core-admin admin traces [OPTIONS]
Inspect autonomous decision traces.
Provides forensics on why AI agents chose specific implementations or strategy pivots.
| Argument / option | Default | Description |
|---|---|---|
--limit, -n |
10 |
Number of recent traces to show. |
--session, -s |
Show specific session details. | |
--failures |
off | Show only traces with violations. |
core-admin admin validate-env
core-admin admin validate-env
Check .env for missing keys or invalid formats. Ensures the 'Body' has correct coordinates for its 'Nerves' (DB/Vectors).
core-admin capabilities
Inspect and search CORE capabilities.
core-admin capabilities search
core-admin capabilities search QUERY [OPTIONS]
Performs a semantic search for capabilities via /v1/search/capabilities.
| Argument / option | Default | Description |
|---|---|---|
QUERY |
required | The semantic query to search for. |
--limit, -n |
5 |
Max results to return. |
core-admin code
Codebase quality, style, and verification operations.
core-admin code audit
core-admin code audit [OPTIONS]
Run the constitutional self-audit.
Default mode: HTTP client over /v1/audit/runs (requires core-api). With --offline: stateless invocation without daemon/DB (F-10 CI gate).
| Argument / option | Default | Description |
|---|---|---|
--severity, -s |
high |
|
--rule, -r |
||
--policy, -p |
||
--files, -f |
Scope per-file rules to these paths (repo-relative or absolute, repeatable). Context-level rules skip with a warning. Closes #279 — enables pre-commit-style focused audits. | |
--verbose, -v |
off | |
--classify |
off | Print context build commands for all findings. |
--force-llm |
off | Bypass the ADR-044 llm_gate verdict cache for this run. Every llm_gate rule re-evaluates every pre-selected file via Ollama; cache rows are still updated with fresh verdicts. Use after suspect cache state or to validate a model upgrade. |
--offline |
off | F-10.1b: run the constitutional audit without core-api / core-daemon / Postgres / Qdrant. Used by the F-10 CI/CD gate and the F-10.5 pre-commit hook. Rules requiring the knowledge graph or LLM (knowledge_gate, llm_gate) are skipped with a structured reason and surfaced in skipped_rules. See ADR-085 §D5 for the F-10 exit criterion this serves. |
--target |
#688: path to an external repo to audit. Requires --offline. The repo must have a .intent/ directory (produced by core-admin project new or core-admin project init). When omitted, the audit runs against the repo found by walking up from the current directory. |
|
--format |
text |
F-10.1b / F-10.2: output format. text (default) renders the human-readable Rich output. json emits the F-10.1a payload verbatim on stdout (schema documented in #535). github-annotations emits GitHub Actions workflow-command lines so findings surface inline in PR diff view (#529 / F-10.2). codeclimate emits CodeClimate JSON for GitLab MR quality reports via artifacts: reports: codequality: (F-10.P2). Non-text formats require --offline; the daemon-driven path has a different result shape. |
core-admin code clarity
core-admin code clarity FILE_PATH [OPTIONS]
Refactor a file for clarity using the V2.3 Adaptive Loop.
This command uses a Cognitive Workflow (Analyze -> Strategize -> Refactor -> Evaluate). It will only apply changes if the Evaluator proves the code is more readable.
| Argument / option | Default | Description |
|---|---|---|
FILE_PATH |
required | Path to the Python file to refactor. |
--write |
off | Actually apply the refactoring to the file. |
core-admin code complexity
core-admin code complexity FILE_PATH [OPTIONS]
Identify and refactor complexity outliers using the V2.3 Adaptive Orchestrator.
Targets high cyclomatic complexity and 'God Methods' to improve modularity.
| Argument / option | Default | Description |
|---|---|---|
FILE_PATH |
required | Path to the file to refactor for complexity. |
--write |
off | Actually apply the refactoring to the file. |
core-admin code refactor-settings
core-admin code refactor-settings [OPTIONS]
Automated refactoring of settings imports to Dependency Injection. Moves layers toward using CoreContext instead of global settings.
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Apply DI refactoring. |
--layers |
mind,will |
Comma-separated layers to process. |
core-admin code verify
core-admin code verify [FILES...]
Judge the current change against the per-file rules and name every rule this run did not evaluate. Exits 1 when the verdict is BLOCKED, 2 when nothing could be judged (NOT_EVALUATED).
Not the authoritative verdict: PASS is given only by code audit.
Example: core-admin code verify
| Argument / option | Default | Description |
|---|---|---|
FILES |
Files to judge. Default: every path that differs from HEAD. |
core-admin cognitive-roles
Cognitive-role capability projection (#821 Unit 2).
core-admin cognitive-roles sync
core-admin cognitive-roles sync [OPTIONS]
Compare cognitive_roles.yaml with core.cognitive_roles; with --write, project the YAML capabilities into the database.
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Write the projection to core.cognitive_roles (default: show drift only). |
--yes, -y |
off | Skip the confirmation prompt for --write. |
core-admin coherence
Constitutional Coherence Checker — scan ADRs, rule domains, and northstar documents for candidate contradictions, gaps, and drift. Governing ADR: ADR-067.
core-admin coherence check
core-admin coherence check [OPTIONS]
Run one Constitutional Coherence Checker pass — 7 check classes per ADR-073.
| Argument / option | Default | Description |
|---|---|---|
--full |
off | Evaluate the full corpus regardless of prior runs. Required for the first run on any installation. Without --full, trigger is auto-detected from the previous run's input_manifest. |
--sample |
Randomly sample N rule files for R2/R3 (R1 still scans all ADRs). Use for narrow exploratory runs. Omit to evaluate all rule files. |
core-admin coherence repair-counts
core-admin coherence repair-counts
Recompute unreviewed_count from live triage state across all open runs.
core-admin coherence report
core-admin coherence report [RUN_ID]
Display the Constitutional Coherence Report for a run.
| Argument / option | Default | Description |
|---|---|---|
RUN_ID |
Run id to report on. Omit to use the most recent run. |
core-admin coherence seed bootstrap
core-admin coherence seed bootstrap
Embed the full governance corpus into the governance_claims collection.
One-shot operation; wall-clock dominated by embedding throughput.
Post-#461 the embed call is batched (~3-10x speedup), so a 1572-claim
corpus completes well under 10 min on the current .40 setup.
The daemon worker handles steady-state incremental sync after this
completes. Run once on install or after a collection drop.
core-admin coherence seed export
core-admin coherence seed export OUTPUT
Dump payloads + vectors from governance_claims to a JSONL fixture.
The fixture restores via coherence seed import without consulting the
embedding endpoint — disaster-recovery path per ADR-073 D4.
| Argument / option | Default | Description |
|---|---|---|
OUTPUT |
required | Destination JSONL path. One line per claim: {payload, vector}. |
core-admin coherence seed import
core-admin coherence seed import SOURCE
Hydrate governance_claims from a JSONL fixture (disaster recovery).
Does not consult the embedding endpoint — vectors come from the fixture.
| Argument / option | Default | Description |
|---|---|---|
SOURCE |
required | JSONL fixture produced by coherence seed export. |
core-admin coherence supersede
core-admin coherence supersede OLD_RUN_ID [OPTIONS]
Retire one CCC run by supersession from another.
| Argument / option | Default | Description |
|---|---|---|
OLD_RUN_ID |
required | Run id to retire (must be currently open). |
--by |
required | Canonical run id that supersedes the old run. |
--note |
required | Mandatory rationale recorded on every dismissed candidate's triage_note. The supersession is auditable through ordinary triage history. |
--yes, -y |
off | Skip the interactive confirmation prompt. |
core-admin coherence triage
core-admin coherence triage CANDIDATE_ID DECISION [OPTIONS]
Record a triage decision on one candidate.
| Argument / option | Default | Description |
|---|---|---|
CANDIDATE_ID |
required | Candidate id to triage. |
DECISION |
required | Triage decision: one of confirmed, dismissed, deferred. |
--note |
Governor rationale. Required when DECISION is 'dismissed'. |
core-admin commands
Search and inspect registered CLI commands.
core-admin commands search
core-admin commands search TERM [OPTIONS]
Fuzzy search across CLI commands via /v1/search/commands.
| Argument / option | Default | Description |
|---|---|---|
TERM |
required | Term to search in command names/descriptions. |
--limit, -l |
25 |
Max results. |
core-admin constitution
Operations for the system mind: policies, schemas, and rule coverage.
core-admin constitution audit
core-admin constitution audit [OPTIONS]
Execute targeted constitutional audits for specific policies or rules.
Example: core-admin constitution audit --policy standard_code_linkage
| Argument / option | Default | Description |
|---|---|---|
--policy, -p |
Policy IDs to execute. | |
--rule, -r |
Specific rule IDs to execute. |
core-admin constitution query
core-admin constitution query TEXT
Perform a semantic search for policies and rules using natural language.
Example: core-admin constitution query "what are the rules for database access?"
| Argument / option | Default | Description |
|---|---|---|
TEXT |
required | Natural language question about the constitution. |
core-admin constitution status
core-admin constitution status
Show enforcement coverage status for the entire constitution.
Identifies 'Enforced' vs 'Declared' rules to detect governance gaps.
core-admin constitution validate
core-admin constitution validate
Validate .intent/ artifacts and .specs/ document headers against their schemas.
MetaValidator checks structural conformance of .intent/ YAML/JSON documents. SpecsDocValidator checks frontmatter headers on modeled .specs/ documents (ADR-105 D6).
core-admin context
Build and explore context packages for LLM assistance.
core-admin context build
core-admin context build [OPTIONS]
| Argument / option | Default | Description |
|---|---|---|
--file, -f |
required | Target file path |
--symbol, -s |
Target symbol name | |
--task, -t |
code_modification |
Task type: code_generation, code_modification, conversational, test.generate, test_generation |
--goal, -g |
`` | Optional goal override |
--show-prompt |
off | Also show assembled LLM prompt |
--output, -o |
Write output to file instead of stdout | |
--no-cache |
off | Bypass context cache |
--max-tokens |
30000 |
Token budget (display only) |
--max-items |
20 |
Max evidence items hint |
core-admin context explain
core-admin context explain QUERY [OPTIONS]
Explore codebase by concept using semantic search.
Answers "what code relates to this concept?" — useful for understanding unfamiliar parts of the codebase before making changes.
For simulating what the agent sees during code generation, use: core-admin context build --file ... --symbol ...
| Argument / option | Default | Description |
|---|---|---|
QUERY |
required | Natural language query |
--output, -o |
Write to file | |
--max-tokens |
30000 |
Token budget |
--max-items |
30 |
Max items |
--no-cache |
off | Disable cache |
core-admin context search
core-admin context search PATTERN [OPTIONS]
Search for code patterns directly in the knowledge graph.
Faster than 'build' for simple pattern searches. Queries the symbols table directly (qualname, signature, intent).
Examples: core-admin context search isinstance core-admin context search "async def" --path src/will core-admin context search "try" --limit 10
For semantic/vector search, use 'core-admin context build' instead.
| Argument / option | Default | Description |
|---|---|---|
PATTERN |
required | Code pattern to search (e.g., 'isinstance') |
--path |
Limit search to specific path | |
--limit |
20 |
Maximum results to return |
core-admin daemon
Background worker daemon management.
core-admin daemon down
core-admin daemon down
Stop all CORE units via systemd.
Covers core-daemon, core-api, and every enabled core-daemon-worker@<stem>.service instance (ADR-081 D6).
core-admin daemon restart
core-admin daemon restart
Restart all CORE units via systemd.
Covers core-daemon, core-api, and every enabled core-daemon-worker@<stem>.service instance (ADR-081 D6).
core-admin daemon start
core-admin daemon start [OPTIONS]
Start background workers (systemd entry point).
Default mode (no --only): discovers daemon-launched active workers
dynamically from .intent/workers/*.yaml — skipping every declaration
with implementation.launch: on_demand (#898) and excluding every
worker that declares requires_dedicated_process: true — and runs the
lightweight set in a shared asyncio loop. This is the unit systemd starts
as core-daemon.
Dedicated mode (--only <stem>): loads exactly that one worker and
runs it alone in its own asyncio loop. The stem must be active,
daemon-launched (not launch: on_demand) and must declare
requires_dedicated_process: true — refuses otherwise. This
is the unit shape systemd starts as core-daemon-worker@<stem>.service
(template unit lands in Step 2c). Per ADR-081 D3 / D5.
Each invocation acquires a singleton PID lock — default mode uses
var/run/core-daemon.pid; dedicated mode uses
var/run/core-daemon-worker-<stem>.pid — so the lightweight daemon
and N dedicated daemons can run simultaneously without contention.
Humans should use core-admin daemon up instead — it goes
through systemctl, returns immediately, and ensures the supervisor
(not your shell) owns the process lifecycle.
| Argument / option | Default | Description |
|---|---|---|
--only |
Run exactly one worker declared requires_dedicated_process: true, alone in its own asyncio loop. Stem is the .intent/workers/<stem>.yaml filename without extension. Without --only, the daemon excludes every worker with requires_dedicated_process: true and runs the remaining lightweight set. Per ADR-081 D5. |
core-admin daemon status
core-admin daemon status
Show CORE service state and scan for stray (orphan) python processes.
Unit list is dynamic per ADR-081 D6 — lightweight core-daemon + core-api plus one row per enabled core-daemon-worker@<stem>.service instance. Drift between .intent/workers/ classification and enabled systemd state is surfaced as a separate panel below the unit table.
core-admin daemon units
core-admin daemon units
Print the systemd units CORE runs, one per line (read by core-deploy, which enables them so they start again after a reboot).
core-admin daemon up
core-admin daemon up
Start all CORE units via systemd (the supported way).
Covers core-daemon, core-api, and every enabled core-daemon-worker@<stem>.service instance (ADR-081 D6).
core-admin database
PostgreSQL state and operational data management.
core-admin database cleanup
core-admin database cleanup [OPTIONS]
Remove orphaned and stale database records.
Cleanup targets: - memory: Old agent episode and reflection entries - action_results: Old records from core.action_results ledger - all: Run all cleanup operations
Constitutional Compliance: - Requires --write flag for safety - Generates audit trail for deletions
Examples: # Dry-run: show what would be deleted core-admin database cleanup
# Clean up old memory entries
core-admin database cleanup --target memory --write
# Full cleanup with custom age threshold
core-admin database cleanup --target all --days 60 --write
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Apply cleanup (default: dry-run) |
--target, -t |
all |
Cleanup target: all, memory, action_results |
--days, -d |
30 |
Age threshold in days for stale data removal |
core-admin database export
core-admin database export
Export database operational data to its canonical YAML representation.
Each operational domain and the vector metadata are written to their fixed, read-only YAML files under the repo (schema-as-truth), via FileHandler. The destination paths are canonical and not configurable.
Constitutional Compliance: - Uses FileHandler for traceable mutations
Examples: core-admin database export
core-admin database migrate
core-admin database migrate [OPTIONS]
Show pending migrations, apply them, or adopt a verified baseline.
Dry run (default): prints pending migration IDs from the manifest. --write: executes pending SQL files, each atomically with its ledger row. --adopt-baseline TAG [--write]: verifies TAG's probes; records ≤ TAG.
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Execute pending migrations (or record the adopted baseline). Without this flag nothing is mutated. |
--apply |
off | DEPRECATED alias for --write; removed next release. |
--adopt-baseline |
Verify that the database is at the declared baseline TAG (e.g. v2.9.1) and, with --write, record the manifest entries through it. Refuses unless every probe of TAG holds and no later baseline also holds. |
core-admin database status
core-admin database status [OPTIONS]
Show database health metrics and diagnostics.
Displays: - Connection status - Database version - Migration status
Examples: # Basic status core-admin database status
# JSON output for scripting
core-admin database status --format json
| Argument / option | Default | Description |
|---|---|---|
--detailed, -d |
off | Show detailed table statistics |
--format, -f |
table |
Output format: table or json |
core-admin database sync
core-admin database sync [OPTIONS]
Synchronize database with codebase symbols.
Scans the codebase and syncs all symbols to the database knowledge graph.
Constitutional Compliance: - Enforces 'knowledge.database_ssot' - All changes logged to audit trail - Atomic action with rollback capability
Examples: # Dry-run (show what would change) core-admin database sync
# Apply changes
core-admin database sync --write
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Apply changes to database (default: dry-run) |
core-admin database sync-registry
core-admin database sync-registry [OPTIONS]
Synchronize the live CLI command structure (metadata) to the database. This makes the DB the Single Source of Truth for available system neurons.
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Actually persist the registry to the database. |
core-admin decisions
Decision history: the repository's ADRs, with their files (ADR-168).
core-admin decisions list
core-admin decisions list [OPTIONS]
List every ADR's id, title and status.
Example: core-admin decisions list --status accepted
| Argument / option | Default | Description |
|---|---|---|
--status |
Only ADRs whose status begins with this, e.g. accepted. |
core-admin decisions show
core-admin decisions show ADR_ID
Show one architecture decision record: id, title, status and its decision headings, with the file it lives in.
Example: core-admin decisions show ADR-168
| Argument / option | Default | Description |
|---|---|---|
ADR_ID |
required | ADR id: ADR-168 or 168. |
core-admin demo
Isolated, opt-in demonstrations that CORE governs its own changes.
core-admin demo cleanup
core-admin demo cleanup RUN_ID [OPTIONS]
Remove (or, without --write, preview) the retained workspace for run_id.
Both modes run the identical marker/containment/symlink/run-id/source-repo
guards via GitService.marker_checked_resolve. Without --write nothing
is removed — the validated target is displayed only. With --write the
same-guarded marker_checked_remove performs the deletion.
| Argument / option | Default | Description |
|---|---|---|
RUN_ID |
required | The run id of the retained demo workspace. |
--write |
off | Actually remove the workspace. Without it, preview the target only. |
core-admin demo consequence-chain
core-admin demo consequence-chain [OPTIONS]
Demonstrate one genuine, isolated governance consequence chain (ADR-155).
Seeds a real violation into a disposable clone, lets the real sensor, remediator, proposal route, executor, and consequence service run it to a committed change and a resolved finding, renders the exact chain as evidence, then tears down all disposable resources — never touching the invoking checkout, its database, or its daemon.
| Argument / option | Default | Description |
|---|---|---|
--output |
Write a Markdown report and JSON companion to this repo-relative path. | |
--keep-workspace |
off | Retain the disposable clone on success (it is always kept on failure). |
--simulate-confirmation |
off | Skip the interactive operator prompt (CI/cold-room). Reported as simulated. |
--timeout-seconds |
Bound the disposable-infrastructure and scenario waits. |
core-admin dev
High-level developer workflows: synchronization, refactor, and stability tools.
core-admin dev analyze
core-admin dev analyze [OPTIONS]
Sense constitutional + structural consequences of a proposed file change.
Builds a shadow workspace where <file> contains <content-from>'s content, runs disk + shadow constitutional audits, runs disk + shadow knowledge graphs, and reports the deltas. Octopus paper Phase 1 — pure sensation, no autonomy, no commit.
| Argument / option | Default | Description |
|---|---|---|
--file, -f |
required | Repo-relative path of the file the crate overlays. Must be under src/ in v1 — other prefixes are guarded against to prevent the shadow materializer writing through directory symlinks into the real repo. |
--content-from, -c |
required | Local file whose contents become the crate-overlaid content for --file. The file on disk is NOT modified; this is purely the proposed-change input the Limb would sense. |
core-admin dev campaign accept
core-admin dev campaign accept CLUSTER_TASK_ID [OPTIONS]
Move an autonomous cluster to 'approved' so it runs on the next execute.
| Argument / option | Default | Description |
|---|---|---|
CLUSTER_TASK_ID |
required | Cluster Task id to accept. |
--write |
off | Apply the acceptance (default dry-run). |
core-admin dev campaign execute
core-admin dev campaign execute PARENT_TASK_ID [OPTIONS]
Run only the clusters the governor has accepted (status='approved').
| Argument / option | Default | Description |
|---|---|---|
PARENT_TASK_ID |
required | Campaign parent Task id. |
--write |
off | Execute approved clusters (default dry-run). |
core-admin dev campaign list
core-admin dev campaign list PARENT_TASK_ID
Show every cluster of a campaign with its id, status, and confidence.
| Argument / option | Default | Description |
|---|---|---|
PARENT_TASK_ID |
required | Campaign parent Task id (printed by strategic-audit --write). |
core-admin dev campaign reject
core-admin dev campaign reject CLUSTER_TASK_ID [OPTIONS]
Move a cluster to 'rejected'; it will not execute. Reason stored in context.
| Argument / option | Default | Description |
|---|---|---|
CLUSTER_TASK_ID |
required | Cluster Task id to reject. |
--reason |
`` | Why the cluster is rejected. |
--write |
off | Apply the rejection (default dry-run). |
core-admin dev campaign stats
core-admin dev campaign stats [CAMPAIGN_ID]
Show accept/reject/pending rates measuring AI-vs-governor divergence.
Without an argument: table of all campaigns with rolling totals. With a campaign id: breakdown for that specific campaign.
High reject rates signal the StrategicAuditor is poorly calibrated for this codebase; low rates confirm its cluster selections align with governor judgment. Per #673 / ADR-110 D4.
| Argument / option | Default | Description |
|---|---|---|
CAMPAIGN_ID |
`` | Campaign parent Task id (omit for rolling stats across all campaigns). |
core-admin dev refactor
core-admin dev refactor GOAL [OPTIONS]
Invoke CORE's autonomous agent loop on a refactoring goal.
CORE will use its own sensory tools (search_vectors, lookup_symbol, read_file) to navigate the codebase and propose a fix — without being handed code directly.
Examples: core-admin dev refactor "Fix logic.logging.standard_only violations" core-admin dev refactor src/shared/utils.py --write core-admin dev refactor "Improve modularity of executor.py" --write
| Argument / option | Default | Description |
|---|---|---|
GOAL |
required | High-level refactoring goal or file path. |
--write |
off | Apply changes. Default is dry-run. |
--workflow |
refactor_modularity |
Workflow type: auto, refactor_modularity, coverage_remediation. |
core-admin dev strategic-audit
core-admin dev strategic-audit [OPTIONS]
Run CORE's strategic self-awareness cycle.
Reads the full system state (audit findings, knowledge graph, constitution, git history), reasons about it as a whole, and produces a prioritised remediation campaign.
Autonomous tasks are staged in PostgreSQL. Tasks requiring .intent/ changes are flagged as escalations for your review.
Examples: core-admin dev strategic-audit # dry-run: report only core-admin dev strategic-audit --write # persist campaign to DB core-admin dev strategic-audit --write --execute # persist + execute autonomously
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Persist campaign to database. |
--execute |
off | Run this campaign's already-approved clusters (implies --write). A new campaign has none yet — review and accept per cluster first via 'dev campaign'. |
core-admin dev sync
core-admin dev sync [OPTIONS]
Run the complete developer synchronization workflow.
- Fixes code (IDs, headers, formatting).
- Syncs knowledge graph to PostgreSQL.
- Syncs semantic vectors to Qdrant.
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Apply all fixes and sync state to DB/Vectors. |
--yes, -y |
off | Skip confirmation prompt for dangerous operations. |
core-admin dev test
core-admin dev test TARGET
Start an interactive test generation session for a specific file.
Allows you to review, edit, and approve test code step-by-step.
| Argument / option | Default | Description |
|---|---|---|
TARGET |
required | Path to the file you want to test. |
core-admin dev test-stability
core-admin dev test-stability ACTION_ID
Runs a stability trial on an action: 1. Runs the action once to apply changes. 2. Snapshots the codebase. 3. Runs the action again. 4. Verifies that the second run resulted in ZERO changes.
| Argument / option | Default | Description |
|---|---|---|
ACTION_ID |
required | The ID of the action to test (e.g. fix.format). |
core-admin docs
Reader-facing documentation generated from live sources (the CLI trees).
core-admin docs generate
core-admin docs generate [OPTIONS]
Regenerate the CLI reference pages and the OpenAPI contract.
Writes docs/reference/core-admin.md (this CLI), docs/reference/core.md (the installed core-cli package) and docs/reference/openapi.json (CORE's public API contract). Runs from a CORE source checkout only, and needs core-cli installed (pip install core-cli).
Example: core-admin docs generate --write
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Write the pages. Without --write, reports which pages would change. |
core-admin grc
Compliance gap-analysis: check a document corpus against a requirements catalog.
core-admin grc gap-analysis
core-admin grc gap-analysis CORPUS [OPTIONS]
Run a compliance requirements catalog against a folder of documents.
Produces a gap report where each requirement is evaluated corpus-wide (ADR-118 D1) and labelled by how its verdict was established: proven (deterministic), judged (AI), or attested (needs a human). Silence from individual documents is not a gap — only corpus-level non-coverage is (D4).
| Argument / option | Default | Description |
|---|---|---|
CORPUS |
required | Folder of documents to analyse (the customer's documentation stack). |
--catalog, -c |
nist_800_171 |
Requirements catalog to check against (a framework under grc-catalogs/). |
--assume-applicable, -y |
off | Skip the applicability confirm step — score even if the framework reads as out of domain for this corpus. |
core-admin grc ingest
core-admin grc ingest FRAMEWORK_ID [OPTIONS]
Ingest a GRC framework's internal corpus into Qdrant (ADR-122).
Enforces the licence gate: copyrighted frameworks require a grc-catalogs/internal/<framework_id>/licence.yaml before ingestion. Public-domain and official-EU-law-reusable frameworks are ungated.
Drops the existing collection and rebuilds from text/ on every run (idempotent re-ingest — safe to run repeatedly after updating source text). Writes a provenance record (licence.yaml) on completion.
| Argument / option | Default | Description |
|---|---|---|
FRAMEWORK_ID |
required | Framework identifier from inventory.yaml (e.g. nist_800_171, gdpr). |
--text-dir |
Directory of section text files to ingest. Defaults to grc-catalogs/internal/<framework_id>/text/. | |
--write |
off | Perform the ingestion (default: dry-run shows what would be ingested). |
core-admin interactive-test
Interactive test generation with step-by-step approval
core-admin interactive-test generate
core-admin interactive-test generate TARGET [OPTIONS]
Generate tests interactively with step-by-step prompts. ... (docstring remains the same)
| Argument / option | Default | Description |
|---|---|---|
TARGET |
required | Module path to generate tests for (e.g., src/shared/models/knowledge.py) |
--write |
off | Actually execute the final creation of the test file. |
core-admin interactive-test info
core-admin interactive-test info
Show information about interactive test generation.
core-admin law
Grounded answers about this repository's law, with sources (ADR-168).
core-admin law access
core-admin law access [OPTIONS]
Every account, group, sudo rule and GitHub collaborator not in the register is a blocking finding (ADR-068 amendment 2026-10-15, rule 5). Exit 1 on any finding.
| Argument / option | Default | Description |
|---|---|---|
--json |
off | Print one JSON line. |
core-admin law apply
core-admin law apply PROPOSAL_ID [OPTIONS]
Approve law proposal PROPOSAL_ID (ADR-170 D3). Run as:
sudo -u core-law /opt/core-law/venv/bin/core-admin law apply <id> --write
Refusals are shown first and nothing is written. Otherwise the warnings follow. Without --write nothing more happens; with --write the proposal is written to every destination, with one signed ledger line, only after you type the proposal id.
| Argument / option | Default | Description |
|---|---|---|
PROPOSAL_ID |
required | Proposal id, e.g. 0020-example. |
--repo |
Development tree (default: discovered from cwd). | |
--key |
core-law's ledger signing key (default ~/.ssh/core_law_ledger). | |
--release-law |
/opt/core/law |
The law beside the release runtime; a second destination when it exists. |
--scratch |
Where the scratch tree is built (default ~/scratch). | |
--write |
off | Ask for the typed id and write. Without it, show the report only. |
core-admin law apply-merge
core-admin law apply-merge PROPOSAL_ID [OPTIONS]
Apply governed text merged on GitHub (one-path flow, 2026-10-12). The
merge into main is the governor's approval; it is recorded as the ledger
approver ("github-merge <sha> by <who>"). Runs only as core-law, started
by root (core-deploy through runuser), never through the governor's sudo
rule, which allows law apply only. Same checks as law apply; a
refusal writes nothing.
| Argument / option | Default | Description |
|---|---|---|
PROPOSAL_ID |
required | The proposal core-deploy built. |
--proposals-dir |
required | |
--merge-commit |
required | |
--merged-by |
required | |
--repo |
required | Development tree. |
--key |
||
--release-law |
/opt/core/law |
|
--write |
off | Apply; without it, report only. |
core-admin law auto-apply
core-admin law auto-apply [OPTIONS]
Apply pending law proposals that change no rule (governor redesign
2026-10-12). A proposal is record-only when it only adds or edits text
under .specs/, removes nothing, and CORE computes no warning for it; it
is applied with ledger approver "auto:record-only" and shows up in
law digest. Every other proposal is listed and left for law apply
with the typed id. Runs only as core-law (its timer, no sudo needed).
| Argument / option | Default | Description |
|---|---|---|
--repo |
Development tree (default: discovered from cwd). | |
--key |
core-law's ledger signing key (default ~/.ssh/core_law_ledger). | |
--release-law |
/opt/core/law |
The law beside the release runtime. |
--write |
off | Apply. Without it, only classify the proposals. |
core-admin law check
core-admin law check PATH
Check whether a producer may write PATH. The answer has two parts that are never collapsed: what the law says, and what actually enforces it (CORE's own write path, and an external assistant editing files directly).
Example: core-admin law check .intent/rules/architecture/governance_basics.json
| Argument / option | Default | Description |
|---|---|---|
PATH |
required | Path to ask about, repo-relative or absolute. |
core-admin law check-tree
core-admin law check-tree [OPTIONS]
Run the D5 refusal checks (schemas, references, the law loading, .claude/settings.json) on the repository in the working directory. Data only: no repository code runs. Exit 1 on any refusal.
| Argument / option | Default | Description |
|---|---|---|
--json |
off | Print one JSON line. |
core-admin law digest
core-admin law digest [OPTIONS]
List law applied in the last HOURS from the ledger, record-only first.
| Argument / option | Default | Description |
|---|---|---|
--hours |
24 |
How far back to look. |
--repo |
Development tree (default: discovered from cwd). |
core-admin law list
core-admin law list [MATCH]
List the rules declared in this repository: id, enforcement, a one-line statement, the engine and check that run it, and the file.
Example: core-admin law list core/starter-python
| Argument / option | Default | Description |
|---|---|---|
MATCH |
Rule-id prefix (starter.) or pack id (core/starter-python). |
core-admin law show
core-admin law show RULE_ID
Show what a rule says: statement, enforcement, authority, and the mechanism that enforces it, with source files. An undeclared rule is answered as unknown.
Example: core-admin law show governance.constitution.read_only
| Argument / option | Default | Description |
|---|---|---|
RULE_ID |
required | Rule id, e.g. governance.constitution.read_only. |
core-admin llm-resources
core.llm_resources authoring/validation surface (#821 Unit 3).
core-admin llm-resources author
core-admin llm-resources author DEFINITION_FILE [OPTIONS]
Validate and, with --apply, persist an llm_resources definition.
| Argument / option | Default | Description |
|---|---|---|
DEFINITION_FILE |
required | Path to a JSON file describing the llm_resources row. |
--apply |
off | Persist the definition to core.llm_resources (otherwise behaves like validate). |
core-admin llm-resources validate
core-admin llm-resources validate DEFINITION_FILE
Validate an llm_resources definition without persisting it.
| Argument / option | Default | Description |
|---|---|---|
DEFINITION_FILE |
required | Path to a JSON file describing the llm_resources row. |
core-admin mcp
The assistant surface over MCP: law, decisions and verdict for any AI assistant (ADR-168).
core-admin mcp run
core-admin mcp run [OPTIONS]
Serve the assistant surface over stdio for one project. Every tool only reads except validate_change and submit_change, which ask a running CORE to check a change and record it as a pending proposal. None can act as the governor. The assistant starts this itself; configuration: https://dariusznewecki.github.io/CORE/connect-an-assistant/
Example: claude mcp add --scope project core -- core-admin mcp run
| Argument / option | Default | Description |
|---|---|---|
--repo |
The project to serve (it, or a parent, must hold .intent/). Default: the working directory. Assistants that do not promise a working directory pass the project root here. |
core-admin project
Operations for project lifecycle: scaffolding.
core-admin project adopt-pack
core-admin project adopt-pack PACK_ID [OPTIONS]
Apply a governance pack to a target repository.
Packs are self-contained bundles of rules and enforcement mappings that remove the need to author governance YAML manually. Packs are resolved from the repository's packs/ registry when it has one (a CORE source checkout), otherwise from the registry bundled with the installed core-runtime.
The target must already carry CORE's machinery floor (.intent/META and the
rest, delivered by project new or project onboard); without it the
command refuses.
Run without --write to preview what would be written. Run with --write to apply. After adoption, run 'core-admin code audit --offline' to see findings against the pack's rules.
| Argument / option | Default | Description |
|---|---|---|
PACK_ID |
required | Pack ID to adopt, e.g. core/starter-python |
--target-dir, -t |
. |
Root of the target repo. Defaults to CWD. |
--write |
off | Apply changes. Without --write, previews what would be written. |
--override |
Downgrade a rule's enforcement. Format: 'rule_id:enforcement', e.g. 'starter.no_bare_except:reporting'. |
core-admin project init
core-admin project init [PATH] [OPTIONS]
Prepare an existing repository to host project law (no law is created).
Copies CORE's machinery floor into <path>/.intent/. The repository is not auditable until a governance pack is adopted or rules are authored.
Example: core-admin project init . --write
| Argument / option | Default | Description |
|---|---|---|
PATH |
. |
The existing repository (default: the current directory). |
--write |
off | Actually copy the machinery floor. |
core-admin project new
core-admin project new NAME [OPTIONS]
Create a new repository ready to host project law (no law is created).
Delivers CORE's machinery floor into .intent/ plus a minimal Python skeleton. The result is not auditable until rules are authored or ratified, or a governance pack is explicitly adopted.
| Argument / option | Default | Description |
|---|---|---|
NAME |
required | The name of the new project. |
--path |
Parent directory for the new project (default: current directory). | |
--write |
off | Actually create the directories and files. |
core-admin refactor
Refactoring analysis and suggestions
core-admin refactor analyze
core-admin refactor analyze FILE_PATH
Analyze a single file for refactoring opportunities.
| Argument / option | Default | Description |
|---|---|---|
FILE_PATH |
required | File to analyze |
core-admin refactor score
core-admin refactor score FILE_PATH
Check the modularity score for a single file with detailed breakdown.
| Argument / option | Default | Description |
|---|---|---|
FILE_PATH |
required | File path to analyze |
core-admin refactor stats
core-admin refactor stats
Show aggregate codebase modularity health using risk tiers.
core-admin refactor suggest
core-admin refactor suggest [OPTIONS]
Rank and suggest files that need refactoring based on current score.
| Argument / option | Default | Description |
|---|---|---|
--min-score |
Filter by score (defaults to constitutional limit) | |
--limit |
50 |
Number of files to show |
core-admin runtime
Runtime state and health of the running CORE system.
core-admin runtime dashboard
core-admin runtime dashboard [OPTIONS]
Five-panel governor dashboard for the CORE runtime.
Shows convergence direction, governor inbox, loop health, pipeline status, and governance coverage — all read-only.
| Argument / option | Default | Description |
|---|---|---|
--plain |
off | Plain text output (log/pipe friendly). |
core-admin runtime health
core-admin runtime health [OPTIONS]
Live health dashboard for the CORE runtime.
Shows workers, observer snapshot, blackboard pulse, recent crawl stats, and blast radius top symbols.
| Argument / option | Default | Description |
|---|---|---|
--plain |
off | Plain text output (log/pipe friendly). |
core-admin runtime sandbox-check
core-admin runtime sandbox-check [OPTIONS]
Run the sandbox closure probes and show which walls held.
| Argument / option | Default | Description |
|---|---|---|
--mode |
service |
service | local |
--socket |
/run/core/sandbox.sock |
The sandbox service's socket. |
core-admin runtime status
core-admin runtime status [OPTIONS]
Last merge, what applied and deployed, and what waits on the governor.
| Argument / option | Default | Description |
|---|---|---|
--status-file |
/var/lib/core-pipeline/status.json |
|
--repo |
DariuszNewecki/CORE |
|
--api |
http://127.0.0.1:8000 |
core-admin secrets
Manage encrypted secrets in the database.
core-admin secrets delete
core-admin secrets delete KEY [OPTIONS]
Delete a secret from the database.
| Argument / option | Default | Description |
|---|---|---|
KEY |
required | Secret key to delete |
--yes, -y |
off | Skip confirmation prompt |
core-admin secrets get
core-admin secrets get KEY [OPTIONS]
Retrieve an encrypted secret from the database.
| Argument / option | Default | Description |
|---|---|---|
KEY |
required | Secret key to retrieve |
--show, -s |
off | Display the secret value |
core-admin secrets list
core-admin secrets list
List all secret keys in the database (does not show values).
core-admin secrets set
core-admin secrets set KEY [OPTIONS]
Store an encrypted secret in the database.
| Argument / option | Default | Description |
|---|---|---|
KEY |
required | Secret key (e.g., 'anthropic.api_key') |
--value, -v |
required | Secret value (will be encrypted) |
--description, -d |
Optional description | |
--force, -f |
off | Overwrite without confirmation |
core-admin status
Single-glance system state and readiness.
core-admin status drift
core-admin status drift [SCOPE]
Consolidated drift entry point.
| Argument / option | Default | Description |
|---|---|---|
SCOPE |
all |
Drift scope: guard|symbol|vector|all |
core-admin symbols
Operations for the symbol registry and Knowledge Graph identification.
core-admin symbols tag
core-admin symbols tag [OPTIONS]
Automatically tag untagged capabilities using AI reasoning.
Consults the TaggerAgent to suggest 'domain.capability' names for symbols that currently have 'unassigned' status in the registry.
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Write suggested tags to the database. |
--limit |
0 |
Limit number of symbols to process (0 = all). |
core-admin tests
Test maturity dashboard and autonomous generation insights.
core-admin tests status
core-admin tests status [OPTIONS]
Single-glance test maturity: sensor health, pipeline stats, coverage gaps.
| Argument / option | Default | Description |
|---|---|---|
--days, -d |
30 |
Lookback window for proposal stats (default 30) |
--gaps, -g |
20 |
Max coverage gaps to show (default 20) |
core-admin tools
Governed, operator-focused maintenance and refactoring tools.
core-admin tools export-context
core-admin tools export-context
Export an operational snapshot into the governed exports directory.
Runs body.maintenance.scripts.context_export.ContextExporter against the CLI's CoreContext. The former --output-dir / --db-url / --qdrant-url / --qdrant-collection options were removed: they were forwarded via a patched sys.argv that the exporter never parsed, and the exporter's coroutine was never awaited, so the command had never run.
core-admin tools rewire-imports
core-admin tools rewire-imports [OPTIONS]
Run after major refactoring to fix all Python import statements across 'src/'.
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Apply the changes to the files. |
core-admin vectors
Vector store operations (Qdrant).
core-admin vectors cleanup
core-admin vectors cleanup [OPTIONS]
Clean up orphaned and inconsistent vector data.
Cleanup targets: - orphans: Vectors in Qdrant without PostgreSQL links - dangling: PostgreSQL links to non-existent vectors - all: Both orphans and dangling links
This performs bidirectional synchronization between Qdrant and PostgreSQL.
Constitutional Compliance: - Enforces 'vectors.bidirectional_sync' - All changes logged to audit trail - Atomic operation with proper ordering
Examples: # Dry-run (show what would be cleaned) core-admin vectors cleanup
# Clean orphaned vectors only
core-admin vectors cleanup --target orphans --write
# Full cleanup
core-admin vectors cleanup --target all --write
| Argument / option | Default | Description |
|---|---|---|
--target, -t |
all |
Cleanup target: 'orphans', 'dangling', 'all' |
--write |
off | Apply cleanup (default: dry-run) |
core-admin vectors rebuild
core-admin vectors rebuild [OPTIONS]
Cold-rebuild the vector store: force re-embed code/tests, then reindex constitutional documents (policies, patterns, specs).
Use after a Qdrant wipe to repopulate all five collections in one operator-invokable command instead of a manual API call sequence.
| Argument / option | Default | Description |
|---|---|---|
--write |
off | Persist rebuilt vectors to Qdrant (default: dry-run). |
--yes, -y |
off | Skip confirmation prompt for this dangerous operation. |
core-admin vectors status
core-admin vectors status
Show vector store health and collection statistics.
core-admin workers
Constitutional worker management.
core-admin workers export-run
core-admin workers export-run RUN_ID [OPTIONS]
Export everything the Blackboard recorded for one run (read-only).
| Argument / option | Default | Description |
|---|---|---|
RUN_ID |
required | The run_id GoalExecutionWorker stamped (goal_run.<run_id>.*). |
--partial |
off | Export a run with no outcome entry, stamped completeness=partial. |
--stdout |
off | Write the export bytes to stdout instead of the canonical file. |
core-admin workers purge
core-admin workers purge [OPTIONS]
Purge blackboard entries by status, with optional rule and age filters.
| Argument / option | Default | Description |
|---|---|---|
--status, -s |
required | Filter by status (required): open | claimed | resolved | abandoned | indeterminate | dry_run_complete | deferred_to_proposal | awaiting_reaudit | suppressed. |
--rule, -r |
Filter by rule ID prefix in subject (e.g. 'style.import_order'). | |
--before, -b |
Only purge entries older than this many hours. | |
--write, -w |
off | Apply deletion. Dry-run by default. |
core-admin workers remediate
core-admin workers remediate [RULE] [OPTIONS]
Run the autonomous remediation pipeline for a constitutional audit rule.
Default (rule mode): sensor finds violations → LLM proposes fix → Canary validates → assisted.validate_diff validates the diff → a passing candidate automatically creates a human-gated PENDING proposal (ADR-154 D3/D5). No fix is applied to src/ or committed until the proposal is approved (currently via the /v1/proposals API — no CLI surface for this yet).
With --sense-only: only the sensor runs — violations are posted to the blackboard but no LLM is invoked.
With --file: run a full audit scoped to a single file across all rules, then feed findings directly into RemediationCeremony (bypasses sensor and blackboard). Mutually exclusive with rule and --sense-only. File-mode findings are synthetic — candidate-export-only (ADR-154 D3a): the validated patch is logged for inspection, never applied, never committed, never submitted as a proposal.
Examples: # Validate a fix and create a PENDING proposal for governor approval core-admin workers remediate purity.no_ast_duplication
# Just sense — populate blackboard, no LLM
core-admin workers remediate purity.no_ast_duplication --sense-only
# File mode — audit one file, remediate all violations found
core-admin workers remediate --file src/body/workers/violation_remediator.py
| Argument / option | Default | Description |
|---|---|---|
RULE |
Audit rule ID to remediate (e.g. 'purity.no_ast_duplication'). | |
--sense-only |
off | Only run the sensor — post findings to blackboard, skip LLM. |
--file |
Audit a single file across all rules, then remediate. Mutually exclusive with rule and --sense-only. Path must be relative to repo root. |
core-admin workers resolve
core-admin workers resolve ENTRY_ID [OPTIONS]
Close a finding delegated to the governor, with a reason.
Acts on 'indeterminate' findings and on 'open' findings whose resolution_mechanism is 'human'. A governor act (ADR-132 D10.4): it goes through POST /v1/findings/{id}/resolve, which only the governor may call, on CORE's Unix socket (CORE_API_URL=unix:///run/core/api/api.sock). The server records who resolved it and under which authority; neither is taken from this command.
| Argument / option | Default | Description |
|---|---|---|
ENTRY_ID |
required | Blackboard entry UUID. |
--reason, -r |
required | Why this finding is being closed. |
core-admin workers run
core-admin workers run WORKER_NAME
Start a constitutional worker by its declaration name.
| Argument / option | Default | Description |
|---|---|---|
WORKER_NAME |
required | Worker declaration name (e.g. doc_worker). |
core-admin workers show
core-admin workers show [OPTIONS]
Inspect the constitutional worker blackboard (read-only).
| Argument / option | Default | Description |
|---|---|---|
--filter, -f |
Filter by subject prefix (e.g. 'ai.prompt.model_required'). | |
--status, -s |
Filter by status: open | claimed | resolved | abandoned | indeterminate. | |
--type, -t |
Filter by entry_type: finding | report | heartbeat | claim | proposal. | |
--limit, -n |
50 |
Maximum number of entries to display. |
--payload, -p |
off | Show full JSON payload for each entry. |