ai-strategy-audit

name	ai-strategy-audit
description	Audits an ai-strategy@1 artefact against ai_strategy_rubric@1.0. Validates FM + SEC structure, EU AI Act risk-classification per use case, NIST AI RMF coverage, MLOps maturity self-assessment honesty, ethics-posture-vs-use-cases internal consistency. Per NIST AI RMF + EU AI Act + ISO/IEC 42001 conventions. Audit one or more existing ai-strategy@1 markdowns against ai-strategy_rubric@1.0. Produces a sibling .audit.md per artefact plus an AUDIT_BATCH_SUMMARY. Halts on needs_human verdicts; resumable on audited_file_sha256. Standalone trigger or chains naturally after ai-strategy-author. Use when user asks to "audit this ai strategy" or "check the ai strategy". Do NOT use for "draft a new ai strategy" (use ai-strategy-author instead). Audit one or more existing ai-strategy@1 markdowns against ai-strategy_rubric@1.0. Produces a sibling .audit.md per artefact plus an AUDIT_BATCH_SUMMARY. Halts on needs_human verdicts; resumable on audited_file_sha256. Standalone trigger or chains naturally after ai-strategy-au...
license	Apache-2.0
metadata	{"version":"1.0.0","module":"skill","stage":"cross","cyberos-template":"ai-strategy@1","cyberos-rubric-version":"ai-strategy_rubric@1.0"}
allowed_memory_scopes	{"read":["project:*","module:*","company:locked-decisions"],"write":["project:*"]}
allowed_mcp_tools	["kb.read","memory.search","audit.append"]
escalation	{"to_persona_on_legal":"cuo-clo","to_persona_on_security":"cuo-cseco","to_persona_on_compliance":"cuo-clo","to_human_on_irreversible":true}
invocation_modes	["standalone","chained"]
expects	{"schema_ref":"./envelopes/input.json","required_fields":["artefact_paths"],"optional_fields":["rubric_version","upstream_context","trace_id","caller_persona","max_iterations_per_artefact"],"standalone_interview_ref":"./STANDALONE_INTERVIEW.md"}
produces	{"schema_ref":"./envelopes/output.json","output_kind":"artefact","human_summary_ref":"./HUMAN_SUMMARY.md"}
depends_on_contracts	[{"id":"ai-strategy","version":"v1","purpose":"validation_target","pin_path":"cyberos/skill/contracts/ai-strategy/"},{"id":"nats-subjects","version":"v1","purpose":"wire_protocol_emission","pin_path":"cyberos/skill/contracts/nats-subjects/"}]
exposable_as	{"internal":true,"agent_plugin":true,"mcp_tool":true,"partner_connector":false}
audit	{"emit_to":"genie.action_log","row_kind":"artefact_write","payload_hash_field":"audited_file_sha256","explanation_pane":"required"}
confidence_band	{"default":0.95,"defer_below":0.5,"cite_sources":"required"}
untrusted_inputs	{"wrap_in_marker":"untrusted_content","injection_scan":"required","on_marker_hit":"surface_to_human"}
self_audit	{"invariants_ref":"./INVARIANTS.md","check_at":["on_node_boundary",{"on_audit_row_count":25},"on_completion"],"anomaly_signals":{"confidence_low_streak":{"threshold":3,"window":10},"user_correction_streak":{"threshold":2,"window":5},"rule_reversal_streak":{"threshold":1,"window":1},"needs_human_rate_above":{"threshold":0.5,"window":10},"deterministic_drift":{"threshold":1,"window":1}},"on_breach":{"emit":"refinement_proposal","pause_pipeline":true,"resume_token_field":"refinement_run_id"}}
human_fine_tune	{"fine_tuner_role":"cpo","review_required":{"on_minor_bump":false,"on_major_bump":true,"on_safety_change":true,"on_rubric_rule_added":true,"on_rubric_rule_removed":true},"signals_to_initiate":[{"acceptance_rate_below":0.6},{"hitl_pause_rate_above":0.4},{"drift_signal_count_above":3},"deterministic_drift_observed","regulator_inquiry_received",{"self_audit_refinement_proposal_count_above":2}],"procedure_ref":null,"required_artifacts":["changelog_entry","acceptance_test_added","rubric_rule_diff","memory_refinement_entry"],"blackout_windows":[]}
determinism	{"reproducible":true,"fixity_notes":"Audit reports are byte-stable for a given artefact + rubric version. Two runs against the same audited_file_sha256 produce identical reports modulo the last_audit_at timestamp."}
emitted_source_freshness_tier	15
gated_until_phase	null
untrusted_content_wrapping	required

ai-strategy-audit — AI_STRATEGY auditor

Standalone trigger that runs ai-strategy_rubric@1.0 against one or more existing ai-strategy@1 markdowns and writes a sibling ai-strategy.audit.md per item. Halts on needs_human verdicts via the standard Question primitive; resumable on audited_file_sha256. Chains naturally after ai-strategy-author.

prompt_revision: ai-strategy_audit@1.0.0

When to invoke this skill

CUO routes a request here when the user wants to:

• "Audit these existing AI_STRATEGYs."
• "Has AI_STRATEGY-007 changed since the last audit?"
• "Tell me which AI_STRATEGYs would fail acceptance today."

Also invoked automatically by the supervisor when ai-strategy-author's output envelope sets next_skill_recommendation: ai-strategy-audit (the default chain).

Self-test preamble

Begin every invocation with a single fenced CONTRACT_ECHO block. Do NOT proceed past this block until it has been emitted.

CONTRACT_ECHO
skill_id:                        ai-strategy-audit
skill_version:                   1.0.0
prompt_revision:                 ai-strategy_audit@1.0.0
template_version:                ai-strategy@1   (loaded from cyberos/skill/contracts/ai-strategy/template.md)
audit_rubric_version:            ai-strategy_rubric@1.0
audit_path_pattern:              <artefact_path with extension replaced by ".audit.md">
hitl_categories:                 [<list per skill>]
hitl_policy:                     HALT_BATCH_ON_NEEDS_HUMAN
max_iterations_per_artefact:     10
re_entrancy:                     idempotent_on_audited_file_sha256
untrusted_content_handling:      spotlight_xml_tagged
file_scope:                      MUST NOT write outside any artefact_path's parent
inputs:
  artefact_paths:                [<list of artefact markdown paths>]
phase:                           AUDIT

§1 Pipeline interface (envelopes)

Input envelope (envelopes/input.json):

{
  "artefact_paths": ["./ai-strategys/AI_STRATEGY-001-foo.md",