ada-doc-implementation-audit

Audits whether project documentation accurately reflects source code, tests, and Git diffs.

Updated Jul 23, 2026
One-click install
npx skills add https://github.com/wubing7755/Ada --skill ada-doc-implementation-audit-wubing7755
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ada-doc-implementation-audit
Source: https://github.com/wubing7755/Ada/tree/main/skills/software-development/ada-doc-implementation-audit
Command: npx skills add https://github.com/wubing7755/Ada --skill ada-doc-implementation-audit-wubing7755

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Project documentation drifts out of sync with the actual codebase: requirement counts go stale, traceability matrices miss REQ IDs, appendix file lists are never updated, and status headers claim work that Git history contradicts. This Skill systematically verifies that SRS documents, traceability matrices, design docs, READMEs, ADRs, and change records tell the truth about what the code actually does. ## Core Features & Use Cases - Full-Project Parallel Audit: Dispatches three parallel subagents to read every source and test file across module clusters (Domain/Interop, Services/Dto, Components), then cross-references findings against all documentation layers to produce a priority-ranked gap report. - Targeted Change-Record Mode: Read-only verification of a single remediation note, fix record, or release note against the current working-tree diff, mapping every checkable claim to file:line and test evidence. - Cross-Reference Automation: Uses grep/comm pipelines to diff REQ IDs between SRS and traceability, compares claimed counts against find/wc/dotnet test output, and classifies gaps into critical, high, and low priority tiers. - Use Case: After a major refactoring phase, ask for a documentation audit to discover that the SRS claims 151 requirements but actually defines 152, the traceability appendix is missing 12 test files, and a design doc marked "pending" was merged weeks ago — then apply fixes in verified batches. ## Quick Start Ask the agent to verify that the project README, SRS, and traceability matrix accurately reflect the current source code and tests, starting with a read-only audit.

Frequently Asked Questions about ada-doc-implementation-audit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I verify that project documentation matches the actual code?

Run a documentation-implementation audit that reads every source and test file, then cross-references claims in the SRS, traceability matrix, README, and design docs against file:line evidence. The audit reports severity-ranked mismatches with a concrete fix plan.

How to check if an SRS traceability matrix is missing requirements?

Extract REQ IDs from both documents with grep, sort them, and run comm -23 to find IDs present in the SRS but absent from traceability. Also run the reverse check, and exclude REQ IDs intentionally marked as deleted before flagging gaps.

Can I audit a single change record against the current Git diff?

Yes, a targeted change-record mode checks one remediation note or release note against the working-tree diff without a full-project audit. It extracts every checkable claim, maps each to current file:line and test evidence, and re-freezes the diff before giving a verdict.

When should I not use a documentation-implementation audit?

Skip it for code-only quality reviews, writing new documentation from scratch, single prose edits with no checkable claims, or pure SRS quality review of terminology and acceptance criteria. Those tasks route to code quality analysis, authoring, or SRS review workflows instead.

Why do documented test counts differ from dotnet test output?

dotnet test counts every Theory data row and MemberData case as a separate test, while appendix counts based on Fact/Theory method declarations are lower. The audit notes this discrepancy explicitly rather than flagging it as an error.

What are the limitations of parallel subagent code audits?

The pattern only works when all source and test files are read; partial audits produce incomplete findings. Subagent output may be truncated, so detailed review must use the full output file path, and timed-out agents should be re-dispatched with smaller file sets.