Skip to content
Security
Command

/org-memory-compile

[experimental] Compile human-reviewed scan history into local organization memory

From plugin
vuln-scout
2215 skills9 agents15 commands
Install
$ npx -y skills add allsmog/vuln-scout --agent claude-code

How it fires

How this command gets triggered: by you, by Claude, or both.

  • Fires itselfClaude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/org-memory-compile

Context preview

What this command does when you run it.

[experimental] Compile human-reviewed scan history into local organization memory

Command definition

org-memory-compile.md
name: org-memory-compile
description: "[experimental] Compile human-reviewed scan history into local organization memory"
argument-hint: "[--privacy open|hashed|strict] [--dry-run] [--allow-commit] [--force]"
allowed-tools:
  - Bash
  - Read
  - Write
  - Glob
  - Grep

Compile Org Memory

Compile repeated human-reviewed verdicts into `.vuln-scout/org-memory/` so future scans can reuse organization-specific suppressions, confirmed patterns, and reviewer heuristics.

Usage

python3 "${CLAUDE_PLUGIN_ROOT}/scripts/org_memory_compiler.py" \
  --project-root . \
  --privacy hashed
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/org_memory_compiler.py" \
  --project-root . \
  --privacy strict \
  --dry-run

Inputs

  • `.claude/scan-history/*.json`
  • `.claude/rule-stats.json`
  • `.claude/review-ledger.json`
  • Human-reviewed findings where `verdict` is `verified` or `false_positive`

Outputs

  • `.vuln-scout/org-memory/accepted-suppressions.yaml`
  • `.vuln-scout/org-memory/confirmed-findings.yaml`
  • `.vuln-scout/org-memory/custom-rules/semgrep/*.yaml`
  • `.vuln-scout/org-memory/review-patterns.yaml`
  • `.vuln-scout/org-memory/manifest.json`

Flags

| Flag | Effect | |------|--------| | `--privacy open` | Store paths and excerpts verbatim | | `--privacy hashed` | Store SHA-256 hashes for paths and excerpts | | `--privacy strict` | Store only rule IDs, verdict counts, and CWE metadata | | `--dry-run` | Print proposed memory without writing files | | `--force` | Allow overwriting strict memory with open privacy | | `--allow-commit` | Do not add `.vuln-scout/org-memory/` to `.gitignore` |

Review Policy

Only human-reviewed provenance can graduate into org memory:

  • Confirmed findings require at least 3 verified samples and a verified rate of at least 50%.
  • Accepted suppressions require the same stable key to be manually suppressed at least twice.
  • Demotion hints reuse `MIN_SAMPLES_FOR_DEMOTE` from `feedback_collector.py` to avoid single-sample overfit.

By default the compiler adds `.vuln-scout/org-memory/` to the target repository `.gitignore`. Use `--allow-commit` only when the repository owner has decided the generated memory is safe to commit.

Read more
Ships withvuln-scout

AI-powered whitebox penetration testing plugin for Claude Code. 9 languages, 22 skills, 7 autonomous agents. STRIDE threat modeling, OWASP 2025 coverage, polyglot monorepo support.

Get the whole plugin, auto-invoked
Stats
22
Stars
1
Views
3
Forks
Maintained
Maintenance
Python
Language
MIT
License
1mo ago
Last commit
5mo ago
Created

Repo: allsmog/vuln-scout