Review documentation problems with AI
Revisit a repeated handover problem using dated evidence. Trace what allowed it to recur, choose a specific repair and decide how the team will check it.
Review documentation problems with AI when a team has fixed the same apparent mistake twice. Perhaps a checklist keeps pointing people at a retired environment. The docs-retro skill can trace the sequence through the accepted instructions, recent edits and earlier audit results before anyone decides that another reminder will help.
A wrong link might be a one-off typo or a symptom of two competing start pages. The remedy depends on the evidence. You get a description of the scope, the proposed change and a test of whether the next reader can complete the task. The exercise can also preserve something that worked and explain when it is worth repeating.
What it does
- Trace why a valid but obsolete guide still appears earlier than its replacement.
- Compare two handovers to establish whether the same navigation failure recurred.
- Identify the missing decision or review step behind conflicting instructions.
- Turn a broad complaint about documentation into one change with a named destination.
- Record the limits of the evidence when a past conversation or version is unavailable.
Install and use
- Open the original docs-governance plugin with an authorised project and relevant records. Python 3 and Git support local checking where needed. Choose a single incident or bounded period.
- Ask the agent to reconstruct what the reader expected, what they encountered and which records corroborate the explanation. Share only pertinent conversation excerpts.
- Pick an improvement and a repeatable test. Authorise any edit explicitly and distinguish the fact that it was made from evidence that it solved the problem.
FAQ
Source
Made by qshanx. Original skill: docs-retro, published on GitHub under the MIT licence.
View on GitHub