metadata-only-wiki-sweep-workflow
Disciplined inventory process for cataloging documents by filename/path without content claims, using parent-centric grouping to prevent stub proliferation
Best use case
metadata-only-wiki-sweep-workflow is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Disciplined inventory process for cataloging documents by filename/path without content claims, using parent-centric grouping to prevent stub proliferation
Teams using metadata-only-wiki-sweep-workflow should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/metadata-only-wiki-sweep-workflow/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How metadata-only-wiki-sweep-workflow Compares
| Feature / Agent | metadata-only-wiki-sweep-workflow | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Disciplined inventory process for cataloging documents by filename/path without content claims, using parent-centric grouping to prevent stub proliferation
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Metadata-Only Wiki Sweep Workflow
Use this when creating stub documentation for large document collections without making content claims. (1) Verify plan approval before execution. (2) Inventory all target directories and extract PDF metadata using `pdfinfo` for safe header-only reading when feasible. (3) Apply parent-centric grouping: merge fragment documents (page scans, figures, sections) into parent document entries rather than creating proliferating stubs. (4) Generate stubs with explicit "do not claim" sections listing what content verification hasn't occurred. (5) Validate stubs with regex checks for prohibited claim language ("normative", "shall", "must", "requires that") in actual content areas, excluding constraint headers.
## Large mixed-directory fallback (learned in Wave 4)
When the collection is very large or noisy (thousands of files with viewer binaries, image bundles, caches, or legacy app payloads), use a conservative fallback instead of forcing a long interactive agent run:
1. Prefer deterministic scripted generation (`execute_code` / Python) over prolonged interactive agent editing.
2. Keep writes inside the approved reporting surface only (for example `docs/reports/**`) if the issue does not require canonical registry/wiki updates.
3. Use conservative file-action classification:
- `stub`: `.pdf`
- `defer`: office/archive/text formats such as `.zip`, `.doc`, `.docx`, `.xls`, `.xlsx`, `.txt`, `.rtf`
- `reject`: viewer binaries, caches, images, shortcuts, and app payloads such as `.dll`, `.fnt`, `.tif`, `.tiff`, `.db`, `.lnk`, `.gif`, `.png`, `.jpg`, `.jpeg`, `.exe`, `.ocx`, `.dat`, `.ini`, `.sys`
4. If PDF header extraction is too expensive for the volume, fall back to filename/path/extension-only metadata and state that explicitly in the artifact header.
5. For high-noise directories, surface the noise in the family map rather than trying to normalize everything into meaningful standards content.
## Interactive-agent recovery rule
If an interactive Codex run drifts into forbidden paths during a metadata-only sweep:
- stop it immediately,
- revert the forbidden path externally,
- narrow the agent back to report-only outputs,
- and if it still stalls, finish the artifact generation programmatically outside the agent session.
Treat the external git state as authoritative; do not trust the agent's verbal claim that cleanup is complete without verifying `git status`.Related Skills
domain-knowledge-sweep
Systematic multi-source research of an engineering domain. Spawns parent issue → 6 research subissues (Standards, Academic, Industry, LinkedIn-marketing, Code-audit, Synthesis) → gap implementation subissues. Replaces LinkedIn-only extraction with defensible comprehensive sourcing.
llm-wiki-weekly-freshness
Class-level governance workflow for keeping llm-wiki-style markdown knowledge bases current, public-safe, graph/index-valid, and useful for code development. Use when reviewing llm-wiki architecture/content, scanning new LLM concepts, maintaining public knowledge graphs, producing an issue roadmap, or running recurring freshness cadence.
llm-wiki-source-extraction-coverage
Doc-type-aware extraction contract for llm-wiki source ingestion with measurable coverage and source-anchored traceability. Use when (1) ingesting a PDF, DOCX, XLSX, PPTX, HTML, or scanned-image source into a wiki `sources/` page, (2) computing the pre-extraction estimate (what fraction of the source we expect to recover) and post-extraction yield (what fraction we actually recovered), (3) anchoring wiki claims back to specific page / paragraph / cell / slide positions in the source so a reviewer can re-verify or revise against the actual document, (4) deciding whether OCR fallback or manual transcription is needed. Codifies workspace-hub's existing OCR fallback chain and python-docx / openpyxl / trafilatura patterns into a format-specific routing table. Companion to research/llm-wiki-page-shape-contract (Rule 7 input-layer pages) and research/llm-wiki — this skill is the defense against silent extraction failure.
llm-wiki-public-private-routing
Firewall between the public llm-wiki repo (vamseeachanta/llm-wiki, MIT + CC-BY-4.0) and per-client private wikis (vamseeachanta/llm-wiki-<client>, e.g. llm-wiki-mkt-a per #2746). Use when (1) deciding whether a converted wiki page lands in public or private surface, (2) applying the project-name abstraction rule to public-bound content, (3) evaluating the public- availability exception that lets actual project names pass through unmodified, (4) promoting content from private to public after sanitization. Encodes the 2026-05-20 user routing directive verbatim: exact client results → private; abstracted (project-name only) → public; project name + all key data publicly available → exception applies. Companion to research/llm-wiki-page-shape-contract (which calls this skill at Rule 8) and research/llm-wiki-source-extraction-coverage (which produces the source pages this skill decides where to send).
llm-wiki-page-shape-contract
Enforce the page-shape contract when a repo-side document or analysis output gets converted into an llm-wiki page. Use when (1) running `scripts/knowledge/llm_wiki.py ingest`, (2) writing or rewriting a wiki page from docs/reports/*, docs/handoffs/*, scripts/review/results/*, or calc citation outputs, (3) deciding whether a page should be split into a folder of sub-pages, (4) reviewing wiki PRs for length / diagram / divide-and-conquer compliance. Codifies the Karpathy + Astro-Han + lewislulu page rules applied to workspace-hub's domain-wiki layout under /mnt/local-analysis/llm-wiki/wikis/<domain>/. Sibling to research/llm-wiki (which owns the CLI ops) — this skill is the quality gate every converted page must clear before commit.
llm-wiki-cadence-governance
Weekly governance workflow for keeping an llm-wiki repository current, code-development-useful, and connected to actionable GitHub issue planning.
llm-wiki-audit-feedback-loop
Durable feedback loop for correcting llm-wiki pages without losing the correction to chat history. Use when (1) a human notices a wiki page is wrong, outdated, or contradicts a source, (2) processing the `audit/` inbox of a domain wiki, (3) reviewing what feedback has been resolved vs deferred, (4) needing to leave a comment on a specific text range that survives line- number drift. Implements the anchored-text audit file pattern from lewislulu/llm-wiki-skill, adapted for workspace-hub's domain-wiki layout under /mnt/local-analysis/llm-wiki/wikis/<domain>/. Extends the 5-op model (compile/ingest/query/lint) from research/llm-wiki with the missing `audit` op. Never silently delete feedback — rejected audits stay archived with rejection rationale.
digitalmodel-orcawave-orcaflex-proof-workflows
Class-level digitalmodel OrcaWave/OrcaFlex readiness, semantic-proof, fixture-proof, and closeout workflows.
oss-wiki-development-arc
Three-phase methodology (Substrate → Depth → Quality) for building open-source engineering wikis efficiently. Skip 70%+ of empirical iteration cost by pre-loading the pattern.
client-llm-wiki-factory
Operator checklist for instantiating a new per-client private llm-wiki repo under workspace-hub [#2746](https://github.com/vamseeachanta/workspace-hub/issues/2746) + [#2731](https://github.com/vamseeachanta/workspace-hub/issues/2731) D4 (amended) naming convention `llm-wiki-<client>`.
read-only-pre-implementation-audit
Systematic cross-check workflow to validate assumptions before TDD coding begins
portable-pattern-verification-workflow
Multi-package implementation with verification strategy for cross-platform configuration hardening