- Home
- Skills
- Vladm3105
- Aidoc Flow Framework
- Doc Adr Reviewer
doc-adr-reviewer_skill
- Python
9
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill vladm3105/aidoc-flow-framework --skill doc-adr-reviewer- SKILL.md16.4 KB
Overview
This skill performs comprehensive content review and quality assurance for Architecture Decision Records (ADR). It checks decision completeness, alignment to BRD topics, consequence and alternative coverage, cross-references, naming, placeholders, and upstream drift. The skill produces a review score and a categorized issue list, and updates a mandatory drift cache used to detect stale upstream documents.
How this skill works
The reviewer loads one or more ADR files, enforces a nested-folder structure, and runs a suite of content checks including decision completeness, BRD mapping, consequence and alternative analysis, cross-reference integrity, naming rules, and placeholder detection. It extracts upstream references (for example @bdd tags and links to ../04_BDD/), computes SHA-256 hashes, compares them against a persistent .drift_cache.json, and flags content drift with severity thresholds. Reports, cache updates, and optional auto-fixes are saved alongside the ADR in the nested folder.
When to use it
- Immediately after ADR generation or autopilot runs to validate content quality
- After manual edits to an ADR to confirm decision rationale and links remain correct
- Before running system-generation or SYS-level automation to ensure ADRs are final
- During formal architecture review sessions as a final QA gate
- As part of periodic maintenance to detect upstream drift and stale decisions
Best practices
- Always store ADRs in the required nested folder format so checks can run
- Run the reviewer after any change to upstream BDD or PRD artifacts to detect drift early
- Address errors marked as blocking (structure and critical drift) before proceeding
- Keep the drift cache under source control or backed up to preserve audit history
- Treat reviewer warnings as actionable items: add mitigations, rationale, or links rather than ignoring them
Example use cases
- Validate that a newly generated ADR includes context, explicit decision, rationale, and date before approval
- Detect that a linked BDD feature changed substantially and flag the ADR for re-evaluation
- Find placeholders like [TODO] or [TBD] left in an ADR and list them for manual update
- Verify BRD Section 7.2 topics are covered and identify missing traceability entries
- Produce a review report with a PASS/WARNING/FAIL score before merging ADRs into the canonical docs
FAQ
The reviewer fails with a mandatory error (REV-D006). The cache must be writable and is created on first run.
Can the reviewer auto-fix content issues?
The tool can apply limited auto-fixes for trivial link or formatting problems, but most content issues are flagged for manual review.