check-source

Files (2)

• SKILL.md
• openai.yaml · agents/

SKILL.md (first lines)

# Check source

Find what looks wrong in one file. Report. Never edit the file.

## Preconditions

Confirm `lstack.yaml` exists in the working directory or a parent. If not, say this isn't a vault and suggest `/setup-vault`. Read `MISSION.md`, `kb/today.md`, and the newest file in `log/`. Open with one line on where the user is.

## Argument

The text after the skill name in the user's message is a file in `sources/` or `notes/`, or a node id. If there is none, ask.

## Steps

1. Read the file. Read every other file in `sources/` that covers the same topic (match by the nodes' `sources` citations and by terms). Read the related nodes' Summaries and Terms.
2. For each suspected error report, in this shape, one block per claim:
   - **Claim**: the sentence or formula, quoted.
   - **Why it looks wrong**: one or two sentences.
   - **What the other source says**: quote and cite, or "no other source covers this; from my own knowledge".
   - **Confidence**: low, medium, or high.
   Report nothing if nothing looks wrong, and say which sources you compared.
3. If the file is in `notes/`, also flag text that would make a good node addition and say `/add-source` or a node edit could take it. Do not write it here.
4. Offer to add each medium or high finding as an `## Open questions` entry on the related node. Preview per node. Wait for yes. Write. Run `node .lstack/lstack.mjs build`, then `node .lstack/lstack.mjs lint`. Report one line. If `lstack.yaml` says `node_available: false`, rebuild `kb/index.md` and `kb/today.md` by hand and say so.

## Preview format

```
### Proposed write 1 of 1: kb/vector-spaces/span.md
**Body**
+ - lecture-2.md says a spanning list is always independent; Axler 2.A gives (1,0),(0,1),(1,1) as a spanning dependent list. Which is right?
```

Reply **yes** to write, **no** to skip, or name what to change.

## Rules

- Never edit anything in `sources/`, `notes/`, or `problems/attempts/`. Never edit generated files.
- Never claim a source is wrong at high confidence from memory alone. Memory-only findings are low or medium.
- No hooks, no sub-agents, no argument substitution.

From sscw32/lstack, MIT.

More from sscw32/lstack