agentsclimarketplace

Llm wiki provenance

Skill po4yka/llm-wiki-skills/skills/llm-wiki-provenance

Inspect and repair evidence links before adding stable IDs. Use when pages have unsupported claims, generated summaries cite other summaries, raw-source links are missing, support labels are wrong, or the user wants a provenance gap report; route deterministic ID insertion to llm-wiki-claim-anchors.From its SKILL.md

Install
npx -y skills add po4yka/llm-wiki-skills --skill llm-wiki-provenance

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its file declares

Copied from the file, not written here

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.3 KB, 602 tokens by cl100k_base, as published. Nobody here has run it

LLM-Wiki Provenance

Goal

Make important wiki claims traceable to raw sources or clearly mark them as unsupported, inferred, or ambiguous.

When to use

  • A page has claims with no visible link back to raw/ sources or source pages.
  • A generated summary cites another generated summary instead of an original source.
  • Support labels (extracted, inferred, synthesis, unsupported) look wrong or are missing.
  • The user asks for a provenance gap report before assigning stable IDs.
  • Current-state or status claims need re-verified evidence rather than an assumed link.
  • Do not use this skill to insert stable IDs; hand that off to llm-wiki-claim-anchors.

Inputs

  • Target pages or folders.
  • raw/ and wiki/ locations.
  • Existing source pages, hashes and frontmatter if present.
  • Desired mode: report-only or patch.

Procedure

1. Select provenance level

Use the smallest useful level:

LevelMeaning
source-levelPage links to the source page or raw source.
claim-levelIndividual claims link to supporting source sections or anchors.
audit-readyClaims include source path, hash, date, model, confidence and review state.

2. Extract claims

For each target page, identify factual claims, recommendations, current-state statements and synthesis claims.

Classify support:

extracted | inferred | ambiguous | synthesis | unsupported | conflicting

Use extracted for claims directly stated in inspected sources. Use synthesis only for editorial conclusions drawn across sources; keep the source links visible.

3. Locate evidence

Search source pages and raw sources. Do not invent evidence. If evidence is not found, mark a gap instead of rewriting the claim.

4. Produce provenance patches

In patch mode, add backlinks near claims using the vault's preferred style. Examples:

- Claim text. Source: [[sources/source-title#section]]

or frontmatter:

source_paths:
  - raw/sources/example.md

5. Create a gap report

Group gaps by severity:

  • high-impact unsupported claims;
  • generated pages citing generated pages;
  • stale current-state claims;
  • inferred claims presented as extracted;
  • ambiguous claims outside review queues.

Output

## Provenance summary

## Pages inspected

## Claims classified

## Provenance added or proposed

## Unsupported claims

## Conflicts and ambiguities

## Suggested next sources

Safety gates

  • Do not fabricate citations or anchors.
  • Do not mark unsupported claims as verified.
  • Do not silently rewrite claims to fit available evidence.
  • Do not expose sensitive source excerpts unnecessarily.
  • If evidence conflicts, hand off to llm-wiki-conflict-resolver.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,144. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.