Skip to content
audit-security logo

Single-Session Security Audit

audit-security

Run a single-session security audit on the codebase

SKILL.md

Full skill instructions

Single-Session Security Audit

Execution Mode Selection

ConditionModeTime
Task tool available + no context pressureParallel~15 min
Task tool unavailableSequential~60 min
Context running low (<20% remaining)Sequential~60 min
User requests sequentialSequential~60 min

Section A: Parallel Architecture (4 Agents)

When to use: Task tool available, sufficient context budget, no S0/​S1 in scope

Agent 1: vulnerability-scanner

Focus Areas:

  • Authentication & Authorization
  • Input Validation & Injection Prevention
  • OWASP Top 10 Coverage

Files:

  • app/​api/​**/​*.ts (API routes)
  • middleware.ts
  • lib/​auth*.ts
  • Form handlers, input components

Agent 2: supply-chain-auditor

Focus Areas:

  • Dependency Security & Supply Chain
  • Package vulnerabilities (npm audit)
  • License compliance
  • Unpinned versions, risky postinstall scripts

Files:

  • package.json, package-lock.json
  • All import statements
  • functions/​package.json

Agent 3: framework-security-auditor

Focus Areas:

  • Firebase/​Firestore Security (rules, Cloud Functions)
  • Next.js/​Framework-Specific Security
  • Hosting & Headers Security
  • Server/​client boundary leaks

Files:

  • firestore.rules, storage.rules
  • firebase.json
  • functions/​src/​**/​*.ts
  • next.config.mjs

Agent 4: ai-code-security-auditor

Focus Areas:

  • AI-Generated Code & Agent Security
  • Crypto & Randomness
  • File Handling Security
  • Product/​UX Security Risks

Files:

  • .claude/ configs
  • Files with crypto, random, hash patterns
  • File upload handlers
  • Admin UI components

Parallel Execution Command

Invoke all 4 agents in a SINGLE Task message:

Task 1: vulnerability-scanner agent - audit auth, input validation, OWASP Task
2: supply-chain-auditor agent - audit dependencies, npm packages Task 3:
framework-security-auditor agent - audit Firebase, Next.js, headers Task 4:
ai-code-security-auditor agent - audit AI patterns, crypto, files

Coordination Rules

  1. Each agent writes findings to separate JSONL section
  2. S0/​S1 findings trigger immediate notification
  3. File conflicts resolved by framework-security-auditor (highest authority)
  4. All agents respect rate-limiting and App Check patterns

Section B: Sequential Fallback (Single Agent)

When to use: Task tool unavailable, context limits, or user preference

Execution Order (priority-first):

  1. AI Security Patterns (S0 potential) - 15 min
  2. Auth & Authorization - 10 min
  3. Input Validation - 10 min
  4. Supply Chain (npm audit) - 5 min
  5. Remaining categories - 20 min

Total: ~60 min (vs ~15 min parallel)

Checkpointing: After each category, write intermediate findings to file before continuing. This protects against context loss.

Checkpoint Format

{
  "started_at": "ISO timestamp",
  "categories_completed": ["Auth", "Input"],
  "current_category": "DataProtection",
  "findings_count": 24,
  "last_file_written": "stage-2-findings.jsonl"
}

Write Checkpoint After Each Category

  • Update ${AUDIT_DIR}/​checkpoint.json
  • Append findings to .jsonl file (not overwrite)
  • This enables recovery regardless of agent capability

Pre-Audit Validation

Step 0: Episodic Memory Search (Session #128)

Before running security audit, search for context from past security sessions:

// Search for past security audit findings
mcp__plugin_episodic -
  memory_episodic -
  memory__search({
    query: ["security audit", "S0", "vulnerability"],
    limit: 5,
  });

// Search for specific vulnerability patterns addressed before
mcp__plugin_episodic -
  memory_episodic -
  memory__search({
    query: ["OWASP", "injection", "auth bypass"],
    limit: 5,
  });

Why this matters:

  • Compare against previous security findings (regression detection)
  • Identify recurring vulnerabilities (may indicate architectural issue)
  • Track which S0/​S1 issues were resolved vs still open
  • Prevent re-flagging known false positives

Step 1: Check Thresholds

Run npm run review:check and report results. Check for security-sensitive file changes.

  • Display count of security-sensitive files changed
  • If none: "⚠️ No security-sensitive changes detected. Proceed anyway?"
  • Continue with audit regardless (user invoked intentionally)

Step 2: Gather Current Baselines

Collect these metrics by running commands:

# Dependency vulnerabilities (extract summary without truncating JSON)
npm audit --json 2>/​dev/​null | node -e '
try {
  const d = JSON.parse(require("fs").readFileSync(0,"utf8"));
  console.log(JSON.stringify(d.metadata?.vulnerabilities ?? d.vulnerabilities ?? {}, null, 2));
} catch (e) {
  console.log("{\"error\": \"Invalid JSON from npm audit\"}");
}
'

# Security lint warnings
npm run lint 2>&1 | grep -i "security" | head -10

# Pattern compliance (security patterns)
npm run patterns:check 2>&1

# Check for .env files (existence only - no permission/​owner metadata needed)
ls .env* 2>/​dev/​null || echo "No .env files found"

Step 2b: Query SonarCloud Security (if MCP available)

If mcp__sonarcloud__get_security_hotspots is available:

  • Query with status: "TO_REVIEW" to get unresolved security hotspots
  • Note hotspot count and severity distribution

If mcp__sonarcloud__get_issues is available:

  • Query with types: "VULNERABILITY" to get security-specific issues
  • Cross-reference with npm audit findings for comprehensive coverage

This provides real-time security issue data from static analysis.

Step 3: Load False Positives Database

Read docs/​audits/​FALSE_POSITIVES.jsonl and filter findings matching:

  • Category: security
  • Expired entries (skip if expires date passed)

Note patterns to exclude from final findings.

Step 4: Check Template Currency

Read docs/​templates/​MULTI_AI_SECURITY_AUDIT_PLAN_TEMPLATE.md and verify:

  • FIREBASE_CHANGE_POLICY.md reference is valid
  • Security-sensitive file list is current
  • OWASP categories are complete
  • Firestore rules path is correct

If outdated, note discrepancies but proceed with current values.


Audit Execution

Focus Areas (12 Categories):

  1. Authentication & Authorization (auth checks, role validation, IDOR, privilege escalation)

  2. Input Validation & Injection Prevention:

    • SQL/​NoSQL injection, command injection
    • Template injection, eval/​Function(), new Function()
    • Unsafe deserialization, prototype pollution
  3. Data Protection (encryption, PII handling, secrets, overly verbose errors)

  4. Firebase/​Firestore Security (rules, Cloud Functions, rate limiting, replay protection)

  5. Dependency Security & Supply Chain:

    • npm audit, outdated packages
    • Unpinned versions, risky postinstall scripts
    • Unused dependencies with known vulnerabilities
  6. OWASP Top 10 Coverage

  7. Hosting & Headers Security:

    • CSP, HSTS, X-Frame-Options, X-Content-Type-Options
    • COOP, COEP, Referrer-Policy, Permissions-Policy
  8. Next.js/​Framework-Specific:

    • Server/​client boundary leaks (secrets in client bundles)
    • API route / middleware auth gates
    • Static export vs server rendering assumptions
  9. File Handling Security:

    • Insecure file upload, path traversal
    • MIME type validation, file size limits
  10. Crypto & Randomness:

    • Weak randomness (Math.random for security), broken hashing
    • Unsafe JWT/​session handling, homegrown crypto
  11. Product/​UX Security Risks:

    • Misleading "security UI" (toggles without server enforcement)
    • Dangerous defaults not clearly communicated
    • Admin-only flows accessible via client routes without backend checks
  12. AI-Generated Code & Agent Security:

    • Prompt-injection surfaces in scripts/​configs
    • Agent config files with unsafe patterns
    • Suspicious strings/​comments that could manipulate AI agents
  13. AI Security Patterns (AI-Codebase Specific):

    • Prompt Injection Surfaces: .claude/ configs, agent prompts, LLM integrations with user input (S0)
    • Hallucinated Security APIs: Security functions that don't exist or have wrong signatures (S1)
    • AI-Suggested Insecure Defaults: Default any types, * CORS, overly permissive rules (S1)
    • Inconsistent Auth Patterns: Different auth approaches across AI sessions (S2)
    • Over-Confident Comments: "This is secure because..." without actual security (S2)
    • AI-Generated Crypto: Homegrown crypto patterns AI might generate (S0)

For each category:

  1. Search relevant files using Grep/​Glob
  2. Identify specific vulnerabilities with file:line references
  3. Classify severity: S0 (Critical) | S1 (High) | S2 (Medium) | S3 (Low)
  4. Classify OWASP category if applicable
  5. Estimate effort: E0 (trivial) | E1 (hours) | E2 (day) | E3 (major)
  6. Assign confidence level (see Evidence Requirements below)

Security-Sensitive Files to Check:

  • firestore.rules, storage.rules
  • functions/​src/​**/​*.ts
  • lib/​firebase*.ts, lib/​auth*.ts
  • middleware.ts, next.config.mjs
  • firebase.json (hosting headers, rewrites)
  • .env* files (environment variables)
  • package.json, package-lock.json (supply chain)
  • .claude/ configs (agent security)
  • Any file with "security", "auth", "token", "secret", "credential" in name

Additional Checks for Vibe-Coded Apps:

  • Search for eval(, new Function(, Function( - dynamic code execution
  • Search for dangerouslySetInnerHTML - XSS vectors
  • Search for NEXT_PUBLIC_ env vars - ensure no secrets leaked
  • Search for process.env in client components - boundary leaks
  • Search for postinstall, preinstall in package.json - supply chain
  • Search for suspicious patterns in .claude/ that could be prompt injection

Scope:

  • Include: app/, components/, lib/, functions/, firestore.rules, firebase.json, .claude/
  • Exclude: node_modules/, .next/, docs/, tests/

Evidence Requirements (MANDATORY)

All findings MUST include:

  1. File:Line Reference - Exact location (e.g., lib/​auth.ts:45)
  2. Code Snippet - The actual vulnerable code (3-5 lines of context)
  3. Verification Method - How you confirmed this is an issue (grep output, tool output)
  4. Standard Reference - CWE number, OWASP category, or security best practice citation

Confidence Levels:

  • HIGH (90%+): Confirmed by external tool (npm audit, ESLint security), verified file exists, code snippet matches
  • MEDIUM (70-89%): Found via pattern search, file verified, but no tool confirmation
  • LOW (<70%): Pattern match only, needs manual verification

S0/​S1 findings require:

  • HIGH or MEDIUM confidence (LOW confidence S0/​S1 must be escalated)
  • Dual-pass verification (re-read the code after initial finding)
  • Cross-reference with npm audit, ESLint, or patterns:check output

Cross-Reference Validation

Before finalizing findings, cross-reference with:

  1. npm audit output - Mark dependency findings as "TOOL_VALIDATED" if npm audit agrees
  2. ESLint security warnings - Mark code findings as "TOOL_VALIDATED" if ESLint flagged same issue
  3. patterns:check output - Mark pattern violations as "TOOL_VALIDATED" if patterns:check flagged
  4. Prior audits - Check docs/​audits/​single-session/​security/ for duplicate findings

Findings without tool validation should note: "cross_ref": "MANUAL_ONLY"


Dual-Pass Verification (S0/​S1 Only)

For all S0 (Critical) and S1 (High) findings:

  1. First Pass: Identify the issue, note file:line and initial evidence
  2. Second Pass: Re-read the actual code in context
    • Verify the vulnerability is exploitable
    • Check for existing mitigations (validation, sanitization, auth checks)
    • Confirm file and line still exist
  3. Decision: Mark as CONFIRMED or DOWNGRADE (with reason)

Document dual-pass result in finding: "verified": "DUAL_PASS_CONFIRMED" or "verified": "DOWNGRADED_TO_S2"


Output Requirements

1. Markdown Summary (display to user):

## Security Audit - [DATE]

### Baselines

- npm audit: X vulnerabilities (Y critical, Z high)
- Security patterns: X violations
- Security-sensitive files: X changed since last audit

### Findings Summary

| Severity | Count | OWASP Category | Confidence  |
| -------- | ----- | -------------- | ----------- |
| S0       | X     | ...            | HIGH/​MEDIUM |
| S1       | X     | ...            | HIGH/​MEDIUM |
| S2       | X     | ...            | ...         |
| S3       | X     | ...            | ...         |

### Critical/​High Findings (Immediate Action)

1. [file:line] - Description (S0/​OWASP-A01) - DUAL_PASS_CONFIRMED
2. ...

### False Positives Filtered

- X findings excluded (matched FALSE_POSITIVES.jsonl patterns)

### Dependency Vulnerabilities

- ...

### Recommendations

- ...

2. JSONL Findings (save to file):

Create file: docs/​audits/​single-session/​security/​audit-[YYYY-MM-DD].jsonl

CRITICAL - Use JSONL_SCHEMA_STANDARD.md format:

{
  "category": "security",
  "title": "Short specific title",
  "fingerprint": "security::path/​to/​file.ts::identifier",
  "severity": "S0|S1|S2|S3",
  "effort": "E0|E1|E2|E3",
  "confidence": 90,
  "files": ["path/​to/​file.ts:123"],
  "why_it_matters": "1-3 sentences explaining security impact",
  "suggested_fix": "Concrete remediation direction",
  "acceptance_tests": ["Array of verification steps"],
  "evidence": ["code snippet", "grep output", "tool output"],
  "owasp_category": "A01|A02|...|A10|N/​A",
  "cvss_estimate": "LOW|MEDIUM|HIGH|CRITICAL"
}

For S0/​S1 findings, ALSO include verification_steps:

{
  "verification_steps": {
    "first_pass": {
      "method": "grep|tool_output|file_read|code_search",
      "evidence_collected": ["initial evidence"]
    },
    "second_pass": {
      "method": "contextual_review|exploitation_test|manual_verification",
      "confirmed": true,
      "notes": "Confirmation notes"
    },
    "tool_confirmation": {
      "tool": "npm_audit|eslint|sonarcloud|patterns_check|NONE",
      "reference": "Tool output or NONE justification"
    }
  }
}

⚠️ REQUIRED FIELDS (per JSONL_SCHEMA_STANDARD.md):

  • category - MUST be security (normalized)
  • fingerprint - Format: <category>::<primary_file>::<identifier>
  • files - Array with file paths (include line as file.ts:123)
  • confidence - Number 0-100 (not string)
  • acceptance_tests - Non-empty array of verification steps

3. Markdown Report (save to file):

Create file: docs/​audits/​single-session/​security/​audit-[YYYY-MM-DD].md

Full markdown report with all findings, baselines, and remediation plan.


Post-Audit Validation

Before finalizing the audit:

  1. Run Validation Script:

    node scripts/​validate-audit.js docs/​audits/​single-session/​security/​audit-[YYYY-MM-DD].jsonl
    
  2. Validation Checks:

    • All findings have required fields
    • No matches in FALSE_POSITIVES.jsonl (or documented override)
    • No duplicate findings
    • All S0/​S1 have HIGH or MEDIUM confidence
    • All S0/​S1 have DUAL_PASS_CONFIRMED or TOOL_VALIDATED
  3. If validation fails:

    • Review flagged findings
    • Fix or document exceptions
    • Re-run validation

Post-Audit

  1. Display summary to user
  2. Confirm files saved to docs/​audits/​single-session/​security/
  3. Run node scripts/​validate-audit.js on the JSONL file
  4. Validate CANON schema (if audit updates CANON files):
    npm run validate:canon
    
    Ensure all CANON files pass validation before committing.
  5. Update AUDIT_TRACKER.md - Add entry to "Security Audits" table:
    • Date: Today's date
    • Session: Current session number from SESSION_CONTEXT.md
    • Commits Covered: Number of commits since last security audit
    • Files Covered: Number of security-sensitive files analyzed
    • Findings: Total count (e.g., "1 S0, 2 S1, 3 S2")
    • Reset Threshold: YES (single-session audits reset that category's threshold)
  6. TDMS Integration (MANDATORY) - Ingest findings to canonical debt store:
    node scripts/​debt/​intake-audit.js docs/​audits/​single-session/​security/​audit-[YYYY-MM-DD].jsonl --source "audit-security-[DATE]"
    
    This assigns DEBT-XXXX IDs and adds to docs/​technical-debt/​MASTER_DEBT.jsonl. See docs/​technical-debt/​PROCEDURE.md for the full TDMS workflow.
  7. If S0/​S1 findings: "⚠️ Critical security issues found. Recommend immediate remediation."
  8. Ask: "Would you like me to fix any of these issues now?"

Threshold System

Category-Specific Thresholds

This audit resets the security category threshold in docs/​AUDIT_TRACKER.md (single-session audits reset their own category; multi-AI audits reset all thresholds). Reset means the commit counter for this category starts counting from zero after this audit.

Security audit triggers (check AUDIT_TRACKER.md):

  • ANY security-sensitive file modified, OR
  • 20+ commits since last security audit

Multi-AI Escalation

After 3 single-session security audits, a full multi-AI Security Audit is recommended. Track this in AUDIT_TRACKER.md "Single audits completed" counter.


Adding New False Positives

If you encounter a pattern that should be excluded from future audits:

node scripts/​add-false-positive.js \
  --pattern "regex-pattern" \
  --category "security" \
  --reason "Explanation of why this is not a security issue" \
  --source "AI_REVIEW_LEARNINGS_LOG.md#review-XXX"

Documentation References

Before running this audit, review:

TDMS Integration (Required)

  • PROCEDURE.md - Full TDMS workflow
  • MASTER_DEBT.jsonl - Canonical debt store
  • Intake command: node scripts/​debt/​intake-audit.js <output.jsonl> --source "audit-security-<date>"

Documentation Standards (Required)

Security (Applicable)

More skills from majiayu000

xiaohongshu logo
majiayu000/claude-arsenal

xiaohongshu

xiaohongshu

286 148
View
agent-task-conductor logo
majiayu000/claude-skill-registry

agent-task-conductor

Conduct multi-agent task orchestration and workflow coordination.

663 1
View
conductor-setup logo
majiayu000/claude-skill-registry

conductor-setup

Initialize project with Conductor artifacts (product definition,

663 1
View
animation-designer logo
majiayu000/claude-skill-registry

animation-designer

Expert in web animations, transitions, and motion design using Framer Motion and CSS

663 1
View
diagramming logo
majiayu000/claude-skill-registry

diagramming

Creates Mermaid and ASCII diagrams for flowcharts, architecture, ERDs, state machines, mindmaps, and more. Use when user mentions diagram, flowchart, mermaid, ASCII diagram, text diagram, terminal diagram, visualize, C4, mindmap, architecture diagram, sequence diagram, ERD, or needs visual docume...

663 1
View
h3-pg logo
majiayu000/claude-skill-registry-data

h3-pg

PostgreSQL bindings for H3 hexagonal grid system. Use when working with H3 cells in Postgres, including spatial indexing, geometry/geography integration, and raster analysis.

23 1
View
conductor-development logo
majiayu000/claude-skill-registry

conductor-development

Context-Driven Development skill for projects using Conductor. Use this skill when you detect a `conductor/` directory in the project, when working on tasks defined in a `plan.md` file, or when the user asks about tracks, specs, or plans. Automatically applies TDD workflow, tracks task completion...

663 1
View
conductor-status logo
majiayu000/claude-skill-registry

conductor-status

Display project status, active tracks, and next actions

663 1
View
dockerization logo
majiayu000/claude-skill-registry

dockerization

Official Stakpak application containerization standard operating procedure, a step-by-step guidline to properly dockerize applications. This is a rule book curated by the Stakpak Team.

663 1
View

Popular AI tools

Kaiber logo
Video

Kaiber

Generate, edit, and beat-sync AI video with leading models in one workspace.

Paid
View
Vimcal logo
Productivity

Vimcal

The world's fastest calendar for remote work

Free
View

Transform Your Design with AI Designer by ImgCreator.ai

Freemium
View
Akool AI logo
Content & writing

Akool AI

Revolutionizing Video Production with AI-Powered Creativity

Paid
View

Extend an image past the frame and let AI fill the new aspect ratio.

Freemium
View
StarByFace logo
Security

StarByFace

Discover your celebrity doppelgänger with StarByFace!

Free
View
C

ChainClarity explains 700+ crypto whitepapers in plain English, with layered summaries, comparisons, research tools, alerts, and a $4.99 Pro plan.

Freemium
View
Opus Clip logo
Coding & apps

Opus Clip

Opus.ai: Revolutionize Your Web Experience

Free
View