Imported from connect2future-org/unified-lms-platform (
.framework-manual/generated-skills/check-against-existing-docs/SKILL.md). Install upstream withnpx skills add connect2future-org/unified-lms-platform --skill check-against-existing-docs. Copyright stays with the author.
Check Against Existing Docs
SOURCE — human-edited. Universal cross-cutting skill.
What this is for
Stops answers from being guessed when the real answer is sitting in a document somewhere. Searches everything, in order of authority, and cites tier + path for every finding so the reader knows how much to trust it.
Authority tiers
| Tier | Authority | What's in it |
|---|---|---|
| 🥇 Gold | Highest | Curated, reviewed project artefacts. The engagement's Terms of Reference is the supreme authority within Gold. |
| 🥈 Silver | Enriched context | Client-supplied clarifications, delivery-team annotations, analysis notes |
| 🥉 Bronze | Raw baseline | Bulk client-supplied documentation, meeting/workshop transcripts |
When tiers conflict, Gold overrides Silver overrides Bronze. Any finding that contradicts the Terms of Reference gets flagged, with the ToR position stated first.
Procedure
- Define the query. What's being investigated? Which deliverable will the answer feed? Full corpus, or a specific subset?
- Search Gold first (start with the ToR), then Silver, then Bronze — never skip a tier unless the question is explicitly scoped to one. Don't stop at the first match; keep looking for corroborating or contradicting evidence.
- Cite every finding with a file path and its tier symbol (🥇/🥈/🥉).
- Flag every conflict explicitly: "Bronze states X (path), but Gold (ToR) states Y (path). Gold position applies."
- State gaps plainly. If nothing was found in any tier, say so — a gap is a finding, not a failure.
- Produce a comparison table if the question spans multiple products or systems, with an explicit confidence rating per row (High/Medium/Low/N/A).
Feeds into
gather-requirements — feed findings into a structured spec. rate-doc-confidence if the search reveals gaps worth scoring formally.