assessment-drift-audit

name	assessment-drift-audit
description	Run the assessment/research drift audit. Surfaces stale assessments, unverified evidence, competing interpretations, superseded research claims, and impacted deliverables using DocGraph's built-in drift engine. Presents findings and optional remediation suggestions without writing research decisions.
triggers	["assessment audit","assessment drift","research drift","research audit","stale assessment","competing interpretations","impacted deliverable"]

Assessment/Research Drift Audit

Invocation — Claude Code: Skill("assessment-drift-audit") | OpenCode / Codex: use triggers above.

Runs the DocGraph assessment drift audit and presents findings for human review. The audit is advisory — findings highlight candidates for action, not authoritative rulings.

Execution

• Model: Haiku (query) → Sonnet (remediation suggestions if requested)
• Effort: Low
• Advisor: Not required
• Core computation: done by DocGraph's GetDriftFindings engine, not by this skill

---

Step 1: Check index status

docgraph_status

Confirm:

• assessment_drift pack is listed under Domain Packs
• Research metadata is indexed (docs with research metadata > 0)

If assessment_drift is not listed, the binary is too old or the pack was never registered — tell the user to upgrade DocGraph and run docgraph index --force.

---

Step 2: Run the drift audit

docgraph_context task="research/assessment drift audit" format=drift_audit

This calls DocGraph's GetDriftFindings engine, which checks:

Code	Meaning
research.stale_assessment	valid_until date has passed; assessment is expired
research.unverified_evidence	last_verified date is older than 180 days (configurable)
research.competing_interpretations	Multiple documents share the same claim_id but have conflicting confidence or analyst_status
research.superseded_claim	A research document with superseded_by set is still referenced by other active documents
research.impacted_deliverable	A deliverable (deliverable_id set) is linked to an expired assessment

---

Step 3: Present findings to the user

Group findings by severity (warnings first, then info).

For each finding:

1. State the file path(s) and finding code
2. Show the message and evidence from the finding
3. Suggest a concrete remediation (below) — but do NOT make the change without confirmation

Remediation patterns

Code	Suggested action
research.stale_assessment	Update valid_until in frontmatter, or add a new superseding assessment
research.unverified_evidence	Re-verify the evidence and update last_verified in frontmatter
research.competing_interpretations	Reconcile the conflicting documents; update confidence and analyst_status to reflect the resolved position
research.superseded_claim	Update citing documents to reference the newer version instead
research.impacted_deliverable	Review the deliverable against the expired assessment; update or note the dependency

---

Step 4: Optional — deep-dive a specific document

To review a specific flagged document in detail:

docgraph_node path="<file-path>"

This shows research metadata, incoming/outgoing references, and section content.

---

Skill boundary

This skill:

• Reads findings from DocGraph's engine (no independent analysis)
• Suggests frontmatter patches for human approval
• Does NOT write research decisions, invent authority, or set confidence/analyst_status without explicit user confirmation
• Does NOT modify any files without user approval

Research authority remains with the document owners and analysts. This skill is a display and triage layer.