Document Finder scans document folders, classifies compliance files, resolves version conflicts, and produces a routing plan for downstream analysis skills.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install document-finder
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install document-finder using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Document Finder?
Document Finder is a compliance document librarian for Openclaw Skills workflows. It does not assess policy quality or compliance posture; instead, it rapidly scans a directory, infers document types from filenames, paths, and minimal content peeks, and then routes each file to the right follow-on skill.
It is built for high-volume triage where precision and speed matter more than deep review. By identifying BAAs, security policies, incident response plans, risk assessments, HIPAA documents, and other compliance artifacts, it creates a clean manifest that separates relevant documents from noise and highlights version clusters that require user confirmation.
Document Finder Use Cases
- Triage a compliance folder before analysis begins.
- Identify BAAs, security policies, IR/contingency plans, risk assessments, and HIPAA documents from filenames and paths.
- Detect version conflicts such as v1/v2, draft/final, archive/backup copies, and ask which version is current.
- Filter out irrelevant files like invoices, resumes, meeting notes, and code/config artifacts.
- Produce a manifest that routes documents to baa-review, hipaa-gap-analysis, framework-mapping, control-assessment, or compliance-qa.
- Scan the current working directory when no path is provided.
How Document Finder Works
- Discover files with Glob using the document pattern and exclude common noise directories such as .git, .venv, pycache, node_modules, and .DS_Store.
- Classify each file from filename and parent-directory signals first, using the compliance signal table to assign a document type with minimal effort.
- Mark unsupported code, config, and environment files as Not Relevant immediately without peeking into their contents.
- For Unknown files only, read the first 800 characters and look for compliance keywords that can confirm a type without full-document analysis.
- Detect version clusters by stripping version markers, draft/final suffixes, date stamps, and archive-directory clues, then pause and ask the user which version is current.
- Build a prioritized manifest that records confidence, classification basis, routing target, and any set-aside notes.
- Output a summary table plus JSON manifest, then stop before any substantive compliance analysis begins.
Document Finder Setup
Document Finder has no external package dependencies, but it does require an agent host that can execute Read, Glob, and Grep.
- Add the skill to your Openclaw Skills-compatible workspace.
- Ensure the runtime exposes the allowed tools listed in the skill header.
- Provide a directory path to scan, or omit the path to use the current working directory.
- Confirm any version clusters before allowing the manifest to finalize.
mkdir -p ~/.openclaw/skills/document-finder
# Place SKILL.md in the directory above
# Example input to the skill
/path/to/compliance-docs
Document Finder Data Schema & Taxonomy
Document Finder organizes its output as a scan manifest plus a routing plan. It reads only enough metadata to classify and route files, and it never produces a substantive compliance assessment.
| Layer | Captured data | Purpose |
|---|---|---|
| Discovery | relative path, last-modified date, extension, skip reason | Build the initial file inventory from Glob results |
| Classification | inferred_type, classification_basis, matched signal, confidence | Assign a document type using filename/path first, then content-peek for Unknowns |
| Version management | cluster members, selected current version, set_aside reason, user choice | Resolve superseded copies before manifesting |
| Routing | recommended_skill, priority, notes | Map each confirmed document to the correct downstream skill |
| Manifest | scan_path, scan_date, total_files_scanned, compliance_relevant, not_relevant, unknown, version_clusters_resolved, documents, set_aside, suggested_next_steps | Final deliverable for the user |
Document type taxonomy
| Type | Typical signals | Recommended downstream skill |
|---|---|---|
| BAA | baa, business associate agreement, dpa, vendor agreement | baa-review |
| Security Policy | security policy, acceptable use, password policy, access control | hipaa-gap-analysis |
| Procedures | procedure, SOP, runbook, job aid | hipaa-gap-analysis |
| IR/Contingency Plan | incident response, BCP, DRP, breach response | hipaa-gap-analysis |
| Risk Assessment | risk assessment, risk register, threat model | hipaa-gap-analysis |
| Vulnerability/Pen Test Report | pen test, vulnerability scan, security audit | control-assessment |
| Compliance Assessment | gap analysis, audit report, POAM | framework-mapping |
| Framework Document | NIST, HITRUST, SOC 2, ISO 27001, CMMC | framework-mapping |
| Technical Security Document | SSP, ACL, diagram, inventory, CMDB | control-assessment |
| Privacy Notice/Policy | privacy notice, privacy policy, PIA | compliance-qa |
| Training Material | training, awareness, attestation | compliance-qa |
| Contract (non-BAA) | MSA, NDA, SLA, SOW | compliance-qa |
| Insurance | cyber insurance, COI, E&O | compliance-qa |
| Vendor Due Diligence | questionnaire, vendor assessment, third party risk | compliance-qa |
| HIPAA Document | HIPAA, PHI, ePHI, 45 CFR | hipaa-gap-analysis |
| Unknown | no clear compliance signal | compliance-qa |
| Not Relevant | invoices, resumes, notes, code/config files | none |
No files are rewritten by the skill; it emits a table for humans and a JSON manifest for automation.
Document Finder Advanced Features
- Filename-first routing that minimizes content reads and keeps the workflow fast.
- Version-cluster detection using version numbers, draft/final markers, date suffixes, and archive-directory patterns.
- Explicit user confirmation for current-version selection, preventing accidental routing of superseded documents.
- Fallback 800-character content peek only for Unknown files, preserving the skill’s routing-only scope.
- Deterministic priority ordering so BAAs and security policies rise to the top of the analysis queue.
- Confidence scoring and classification-basis tracking for auditability and downstream automation.
- Built-in routing map to specialized follow-on skills such as baa-review, hipaa-gap-analysis, framework-mapping, control-assessment, and compliance-qa.
- Openclaw Skills-friendly manifest output that can be handed directly to other agents or batch workflows.
SKILL.md
Loading
Related Openclaw Skills
multi-clawd
multi-clawd pools Claude Max accounts into one OpenClaw failover chain so the same model keeps running as accounts rotate.
API Rate Limit Advisor
Openclaw Skills API Rate Limit Advisor helps teams design resilient throttling, quota, and retry strategies for APIs.
NotebookLM CLI Automation
Openclaw Skills turns NotebookLM into a terminal-first workflow for notebooks, source ingestion, Q&A, summaries, and study asset generation.
Prompt Fidelity Self-Check
Openclaw Skills Prompt Fidelity Self-Check quantifies how much of a response is verifiable versus subjective, then reports the audit split and confidence framing.
Astro-ph Radar
Astro-ph Radar is an Openclaw Skills runbook that turns an astrophysics topic into a fresh arXiv scan with full abstracts, authors, dates, and PDF links.
Ben McAdams MCP
Openclaw Skills connects AI agents to Ben McAdams's public MCP server for structured access to his resume, portfolio, and blog notes.
Featured*
