agentleFS
anh-chuanh-chu/llm-wiki-pm/skills/llm-wiki-brief/SKILL.md

llm-wiki-brief

Skill7 starsChanged 9 days agoMITView on GitHub

Structured daily/weekly briefs, tag digests, and check-ins from the PM wiki. Richer output than core catch-me-up.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read $WIKI/_status.md if it exists — surface any pre-computed warnings immediately

    line 33
  2. Very specific:

    Read last 20-30 lines of $WIKI/log.md — know what's recent before summarizing

    line 34
  3. Very specific:

    Run ≥2 varied queries per source before calling it empty (broaden terms, widen the window, try aliases/handles/scope variants, switch …

    line 45
  4. Very specific:

    Before you report findings or a brief, run the llm-wiki-review sub-skill on it: a separate reviewer checks every statement against its …

    line 61
  5. Distinctive:

    If 24h window is empty, expand to 48h and note it.

    line 100
  6. Distinctive:

    Log: Read-only — do not append to log.md unless user asks to file output.

    line 102

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

LLM Wiki Brief

Structured briefs, digests, and check-ins from the PM wiki. Richer, more structured output than the §10/§11 fallbacks in the core llm-wiki-pm skill.

For ingest, update, query, lint, and all other wiki operations — the core llm-wiki-pm skill handles those. This skill is brief-only.

When This Skill Activates

  • "daily brief", "what happened today", "what happened yesterday"
  • "weekly brief", "weekly checkin", "what happened this week", "week in review", "catch me up on the week"
  • "[tag] digest", "brief me on [tag]", "summarize [tag] pages", "what's new in [topic]", "what changed in [competitive/customer/strategy/etc]"
  • "what's thin", "what needs more work", "coverage brief"

Wiki Path Resolution

WIKI=$(cat .wiki-path 2>/dev/null | tr -d '[:space:]')
CONF='${user_config.wiki_path}'; case "$CONF" in '${'*) CONF= ;; esac  # plugin option, if set
WIKI=${WIKI:-${CONF:-${WIKI_PATH:-}}}
[ -z "$WIKI" ] && [ -f SCHEMA.md ] && WIKI=$(pwd)  # cwd only if it is already a wiki

Orient (lightweight)

This skill is read-only by default. Before any brief:

  1. Read $WIKI/_status.md if it exists — surface any pre-computed warnings immediately
  2. Read last 20-30 lines of $WIKI/log.md — know what's recent before summarizing

Full orient (AGENTS.md 4-step protocol) only required if filing output as a wiki page.

Source-Completeness Guard (when briefs sweep live sources)

When a brief is asked to pull from external sources (whatever the user's connected stack — issue trackers, mail, chat, CRM, meeting notes, web), the core llm-wiki-pm Source-completeness guard applies in full: an empty or errored source is a signal to vary the query, not to declare the sweep done.

A brief is exec-facing synthesis — the Provenance-tier & falsification guard applies (core skill → Session Defaults). A brief reads as a complete, current picture, so a laundered claim does the most damage here. Tier every headline claim computed/primary/recalled; don't state a claim recalled from wiki prose as current fact without confirming it against a live source (a wiki page saying "pipeline is thin" does not override what the CRM says today). For any decision-bearing line, name and run the query that would disprove it before it goes in the brief, and mark claims you couldn't verify as unverified rather than asserting them.

  • Get the source set from the registry, not from memory. Before sweeping, read the ## Sweep Registry in $WIKI/MY-INTEGRATIONS.md and cover every source listed; the ledger below must have one row per registered source. If no registry exists, build the set from connected tools + the Active Sources table, confirm it with the user, and write it to the registry. Forgetting a registered source (e.g. omitting Slack or GitHub from the sweep) is the worst failure — it reads as complete while a high-signal source went unread.
  • Run ≥2 varied queries per source before calling it empty (broaden terms, widen the window, try aliases/handles/scope variants, switch syntax). Keyword search tools commonly miss on first phrasing.
  • Classify each requested source as hits / empty-after-retries / failed. Never collapse failed or empty-after-retries into "covered."
  • Always emit a Source Coverage ledger (see below) listing every requested source and its state. A source that returned nothing must appear in the brief, not vanish from it. No "enough captured" while a requested source is failed or was abandoned after one query.
### Source Coverage
- <source A>: hits (N items)
- <source B>: hits (N threads)
- <source C>: empty-after-retries — tried "X", "<scoped query> Y", "<author filter> Z"; quiet or query miss. Confirm scope/range?
- <source D>: failed — not connected

Review before reporting

Before you report findings or a brief, run the llm-wiki-review sub-skill on it: a separate reviewer checks every statement against its sources and searches for what is missing; you revise and end with the "Reviewed by" line.

Operations

1. Daily Brief

Triggers: "daily brief", "what happened today", "what happened yesterday"

① Resolve $WIKI. Read $WIKI/_status.md if present.

② Read $WIKI/log.md. Filter entries from the last 24-48 hours. Parse:

  • Pages created (ingest, create)
  • Pages updated (update)
  • Decisions logged (decision)
  • Open questions added (query, question)

③ For decision and question entries, read the referenced page to pull the one-line summary or question text.

④ Output format — structured, scannable, no prose padding:

## Daily Brief — YYYY-MM-DD

### New Pages
- [[slug]] — one-line description

### Updated Pages
- [[slug]] — what changed (one phrase)

### Decisions Logged
- [[slug]]: decision text (YYYY-MM-DD)

### Open Questions Added
- [[slug]]: question text

### Warnings
- (staleness, coverage flags, lock state from _status.md)

If 24h window is empty, expand to 48h and note it. If still empty, offer weekly brief.

Log: Read-only — do not append to log.md unless user asks to file output.

If you file the brief as a page (e.g. briefings/YYYY-MM-DD.md), stamp lifecycle: dated-digest in its frontmatter. These are orphans by design; the flag keeps them out of the lint orphan count (see schema-guide). Without it, every daily brief permanently inflates the orphan metric.


2. Weekly Brief / Check-in

Triggers: "weekly brief", "weekly checkin", "what happened this week", "week in review", "what did I miss"

① Resolve $WIKI. Read $WIKI/_status.md if present.

② Read $WIKI/log.md. Filter entries from last 7 days (default; user can specify N days).

③ Aggregate counts: pages created, updated, decisions logged, questions added vs resolved.

④ Identify top 3 areas of activity: grep mentioned slugs' frontmatter tags, count by tag frequency.

⑤ Lightweight staleness check on overview.md and index.md: if updated: older than 14 days AND log shows writes this week, surface warning.

⑥ Coverage gaps: grep for coverage: stub pages created this week only — not all stubs, just new ones.

⑦ Output format:

## Weekly Brief — Week of YYYY-MM-DD

### Summary
- Pages created: N | updated: N | decisions: N | open questions: N added, N resolved

### Top Activity Areas
1. [tag] — N entries, top pages: [[a]], [[b]]
2. [tag] — N entries, top pages: [[c]]
3. [tag] — N entries, top pages: [[d]]

### Decisions This Week
- [[slug]]: decision text (YYYY-MM-DD)

### Open Questions Added
- [[slug]]: question text

### Staleness Flags
- [[overview.md]] last updated YYYY-MM-DD — may need refresh

### Coverage Gaps from This Week
- [[slug]] created as stub — gaps: X, Y

### Warnings
- (from _status.md)

⑧ File as queries/weekly-brief-YYYY-MM-DD.md after a full orient per AGENTS.md, and say so.

Log: ## [YYYY-MM-DD] brief | weekly | last 7 days | filed: yes/no — only if filing.


3. Tag Digest

Triggers: "[tag] digest", "brief me on [tag]", "summarize [tag] pages", "what's new in [tag]"

① Resolve $WIKI. Check tag exists in $WIKI/SCHEMA.md taxonomy. If not: "Tag '[x]' not in taxonomy. Did you mean: [closest matches]?"

② Grep for matching pages:

grep -rl "tags:.*\b<tag>\b" $WIKI/entities $WIKI/concepts $WIKI/comparisons --include="*.md"

③ Sort by updated: desc. Read each page: title, type, updated, coverage, confidence, 2-3 sentence summary.

④ Cross-reference with log.md: grep each slug in last 30 days. Note recent activity.

⑤ Identify themes: recurring entities/concepts, what changed recently, open questions (question-tagged pages with matching tag), patterns, gaps.

⑥ Output format:

## [Tag] Digest — YYYY-MM-DD
N pages tagged [tag]

### Key Themes
- [Theme]: [2-3 sentence synthesis]
  Sources: [[page-a]], [[page-b]]

### Recently Updated (last 14 days)
- [[slug]] (YYYY-MM-DD): what changed

### Open Questions
- [[query-slug]]: question text

### Notable Patterns
- [Pattern observed across pages]

### Coverage Notes
- [[slug]]: stub — gaps: X, Y

⑦ File as queries/[tag]-digest-YYYY-MM-DD.md after a full orient, and say so.

Log: ## [YYYY-MM-DD] digest | tag: <tag> | N pages | filed: yes/no — only if filing.


4. Coverage Brief

Triggers: "what's thin", "what needs more work", "coverage brief"

① Scan coverage: frontmatter across entities/, concepts/, comparisons/. Count: stub, partial, comprehensive, missing.

② For stub/partial pages, collect tags: and gaps:. Group by primary tag.

③ Output format:

## Coverage Brief — YYYY-MM-DD

### Summary
- Stub: N | Partial: N | Comprehensive: N | No marker: N

### Stubs by Domain
**[competitive]** — N stubs
- [[slug-a]], [[slug-b]]
- Top gaps: X, Y

**[customer]** — N stubs
- [[slug-c]]

### Partial Pages with Notable Gaps
- [[slug]]: gaps: [list]

### Recommended Next Actions
- [[slug]]: highest-priority — appears in N cross-references

④ Offer: "For full audit with orphan detection and taxonomy analysis, run Coverage Audit (§12) from llm-wiki-pm."

No log entry — pure read operation.


Boundaries

This skill handles: daily brief, weekly brief, tag digest, coverage brief.

For everything else — ingest, update, query, lint, archive, pre-meeting briefing — the core llm-wiki-pm skill handles those.

"For full ingest/update/query operations, the core llm-wiki-pm skill handles those."

Behavioral Constraints (per AGENTS.md)

  • No writes without orient. Filing a brief requires full orient + SCHEMA.md compliance.
  • No schema drift. Filed query pages use taxonomy tags only.
  • No orphan pages. Filed briefs need min 2 outbound [[wikilinks]].
  • Privacy. Pages are private by default (only shareable: true pages are export-safe). For any non-shareable page, surface topic category only, not content.
  • Human tone. Scannable over verbose. No em-dashes. No AI tells.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.