strataudit

name	strataudit
description	Use when the user needs evidence-backed alignment analysis, traceability gaps, coverage, or a source-grounded evidence pack across strategy, architecture, design, or implementation documents in a real corpus.
version	1.1.0

@strataudit - Strategy Traceability Audit

Run a document-backed strategy audit over a real corpus or existing .strataudit artifacts. This skill is for evidence-backed audit work, not free-form strategy prose.

Use When

• the user needs alignment analysis across strategic and delivery documents
• the answer must be backed by extracted entities, traces, findings, or saved artifacts
• the user needs one of these modes: corpus-audit, traceability-audit, coverage-audit, evidence-pack, report-redraft

Do Not Use When

• the user wants only a short narrative summary
• there is no accessible corpus and no existing artifacts
• the task is brainstorming or roadmap generation from scratch
• the problem is operational debugging rather than document traceability

Safety Guards

• only rely on real document text or saved audit artifacts
• do not fabricate quotes, traces, or initiatives
• preserve source language in the evidence layer unless explicitly asked to derive display text
• similarity alone is never enough to call a trace verified
• if provenance is weak, downgrade the claim or refuse to make it

Audit Modes

Mode	Use when	Must emit
corpus-audit	corpus quality and ingest readiness are unclear	corpus inventory, exclusions, level coverage, caveats
traceability-audit	the user wants cross-level alignment and missing links	traces, findings, caveats
coverage-audit	the user asks what is and is not covered	coverage summary with explicit denominators
evidence-pack	the user wants inspectable proof behind claims	source-backed findings, trace tables, caveats
report-redraft	the user wants a better report from existing artifacts	rewritten sections with unchanged trust boundaries

Start with corpus-audit when corpus quality is unknown. Use report-redraft only when an evidence pack or prior audit artifacts already exist.

Runtime Order

1. use an injected host-native runtime from the harness when available
2. otherwise use a configured OpenAI-compatible runtime
3. use OpenRouter as the default network enhancer/fallback
4. use artifact-only mode when the question can be answered from existing outputs
5. use sdp-strataudit run only as CLI fallback

The CLI can resolve configured network runtimes. It cannot create a host-native runtime on its own.

Workflow

1. choose the audit mode that matches the user's question
2. validate inputs and trust boundary
3. resolve runtime by policy or reuse existing artifacts
4. run ingest → extract → link → analyze → report, or inspect saved artifacts
5. return artifact paths plus trust caveats and what is not claimed

Refuse When

• the user asks for verified alignment without inspectable provenance
• the corpus root is missing and there are no prior artifacts
• the requested mode needs runtime capabilities that are unavailable
• the requested summary is broader than the evidence pack supports

Output

• .strataudit/report.json
• .strataudit/report.html
• .strataudit/similarity_distribution.json
• .strataudit/strataudit.db
• explicit runtime choice or artifact-only path
• key trust caveats and what is not claimed

References

• docs/QUICKSTART.md
• docs/reference/skills.md
• docs/reference/strataudit-evidence-policy.md
• docs/reference/strataudit-runtime-policy.md
• docs/reference/strataudit-output-modes.md
• sdp-strataudit run

Recovery

Symptom	Fix
Skill produces no output	Check working directory is project root with docs/workstreams/backlog/
"checkpoint not found"	Run sdp-orchestrate --feature <ID> to create initial checkpoint
"workstream files missing"	Run sdp-orchestrate --index to verify, then @feature to regenerate
Skill hangs / no progress	Check .sdp/log/events.jsonl for last event; use sdp reset --feature <ID> if stuck
Review loop exceeds 3 rounds	Use @review --override "reason", @review --partial, or @review --escalate