library-manager
Skill1 starsChanged 7 months agoApache-2.0View on GitHub
Manage review library assets: writing samples, style profiles, checklists, and known-issue patterns. Good samples for style fingerprinting should be: - Final versions (not drafts or work-in-progress) - Same document type as the documents under review (e.g., advisory opinions for advisory review) - Same originating agent if comparing against a specific agent's output - At least 3 pages in length (short documents don't provide reliable style metrics) - Free of known critical issues (samples with hallucinated citations skew metrics) - Representative…
In this repository15 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- citation-auditorSkill
- citation-checkerSkill
- cover-memo-writerSkill
- cross-document-checkerSkill
- document-parserSkill
- formatting-reviewerSkill
- ingestSkill
- known-issues-managerSkill
- quality-gateSkill
- redline-generatorSkill
- scoring-engineSkill
- structure-checkerSkill
- substance-reviewerSkill
- writing-quality-reviewerSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 16
After ingestion, report current sample count and whether style profile can be (re)generated
- Very specific: line 20
Run build-style-profile.py to aggregate metrics across all samples
- Distinctive: line 73
Same originating agent: if comparing against a specific agent's output
≈ “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.
Files
library-manager Skill
Manage review library assets: writing samples, style profiles, checklists, and known-issue patterns.
Capabilities
- Library Status (
/library list)- Scan all library directories and report counts + status
- Output: asset summary table (samples, checklists, known-issues, style-profiles)
- Sample Ingestion (
/library add-sample)- Accept DOCX, PDF, or Markdown files
- Extract text using
document-parserscripts or plain read - Compute style metrics using
ingest-sample.py - Store extracted text + metrics in
library/samples/{filename}.json - After ingestion, report current sample count and whether style profile can be (re)generated
- Style Profile Generation (
/library style-profile regenerate)- Requires ≥5 samples in
library/samples/ - Run
build-style-profile.pyto aggregate metrics across all samples - Compute: mean + standard deviation for each metric (avg_sentence_length, passive_voice_ratio, formality_score, avg_paragraph_length, citation_density)
- Output:
library/style-profiles/{profile_name}.json /library style-profile view— display current profile metrics
- Requires ≥5 samples in
- Checklist Management (
/library add-checklist)- Validate YAML structure against expected schema (document_type, language, description, dimensions with items)
- Save to
library/checklists/{document_type}-{language}.yaml /library add-checklistwith no file → interactive: ask document type, language, then generate skeleton
- Known Issues Management (
/library known-issues)list— display all patterns grouped by agent, with frequency and last_seenadd— interactive: collect dimension, pattern, detection_rule, recommended_fix; assign pattern_idedit {pattern_id}— modify an existing patterndelete {pattern_id}— remove pattern (with confirmation)
Scripts
scripts/ingest-sample.py
Extract text from a file and compute style metrics for library storage.
- Input: file path (DOCX, PDF, MD, or TXT)
- For DOCX: delegates to
document-parser/scripts/parse-docx-structure.pythen reads full_text - For others: direct text read
- Computes metrics using same logic as
style-fingerprint-compare.py - Output: JSON to stdout with
{filename, source_path, ingested_at, text_length, metrics} - Usage:
python3 ingest-sample.py <file_path>
scripts/build-style-profile.py
Aggregate all sample metrics into a style fingerprint profile.
- Input: samples directory path, output path
- Reads all
*.jsonfiles in samples directory - Computes mean and standard deviation for each metric
- Output: style profile JSON with
{profile_name, sample_count, created_at, metrics (means), standard_deviations} - Usage:
python3 build-style-profile.py <samples_dir> <output_path>
scripts/validate-checklist.py
Validate a YAML checklist file against the expected schema.
- Input: YAML file path
- Checks: required top-level keys (document_type, language, description, dimensions), dimensions is a dict with list values
- Output: JSON with validation result + any errors
- Usage:
python3 validate-checklist.py <yaml_path>
When to Use
- WF4:
/librarycommand and all sub-commands - Post-review: when
known-issues-managerproposes a new pattern - Onboarding: when setting up library for a new document author/agent
Sample Quality Guidelines
Good samples for style fingerprinting should be:
- Final versions (not drafts or work-in-progress)
- Same document type as the documents under review (e.g., advisory opinions for advisory review)
- Same originating agent if comparing against a specific agent's output
- At least 3 pages in length (short documents don't provide reliable style metrics)
- Free of known critical issues (samples with hallucinated citations skew metrics)
- Representative of the agent's typical output (not edge cases or unusual formats)
Checkpoint
Library management operations (WF4) do not use the checkpoint system — they are atomic operations that complete immediately.
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
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.

