Security Review
Purpose
Find real, exploitable security weaknesses and report them safely. Implements ../../system/SECURITY_RULES.md. Supports Gate 6 and any security-audit request.
When to Use
- Before merge/release for changes touching auth, data, payments, or external input.
- For a security audit request type.
- Whenever an audit or review surfaces a security-relevant surface.
Inputs
- The change/codebase in scope and how it handles secrets, auth, data.
- Whether the app handles money or sensitive PII.
Discovery Questions
- Does the change touch secrets, auth, authorization, input handling, PII, or payments?
- Is this pre-launch, live, or post-incident?
- What is the trust boundary and the data sensitivity?
Responsibilities
Assess, per ../../system/SECURITY_RULES.md:
- Secrets (no hardcoding; git history; client-exposed prefixes; correct key placement).
- Auth (enforced routes, JWT/session integrity, reset flow, rate limits, no default admin).
- Authorization/IDOR (ownership checks, server-side roles, tenant scoping).
- Input/injection (parameterized queries, XSS, command, uploads, validation).
- PII (minimized in logs/third-party; deletion path).
- Transport/errors (TLS, CORS, security headers, generic errors).
- Payments (server-side totals, webhook signatures, abuse) when applicable.
- Production hardening (debug/test endpoints, env fail-fast, dependency advisories).
Report findings with severity and Confirmed vs Potential.
Required Workflow
- Scope to the change/surface + data sensitivity.
- Walk the relevant security areas. Check that security behavior is specified as scenarios — cross-tenant and cross-user denial, expired or invalid tokens, role escalation refused (
../../system/GHERKIN_RULES.md) — and that each has a passing test. A scenario states the expectation; only tests and review are evidence. Never conclude "secure".
- Run available tools where possible; quote or mark "unverified until run."
- Record findings (severity, Confirmed/Potential,
file:line, redacted).
- List secrets to rotate; recommend human review for money/PII at scale.
- On a tracked project, record them via
project-management: each Confirmed finding through bug intake (Critical/High → P0/P1), Potential ones as TECH- tasks to verify (../../system/PROJECT_MANAGEMENT_RULES.md, recording work).
Decision Rules
- Never print a discovered secret's value — report location/type, redacted, and flag for rotation.
- Separate Confirmed from Potential; assign severity (Critical/High/Medium/Low).
- Block release on open Critical/High confirmed issues.
- Never claim "secure"; report what was checked and fixed, and what needs human review.
Rules
- Do no harm: no data exfiltration, no hitting production/third-party systems, no destructive commands.
- Do not weaken existing controls to "make it work."
- Recommend a professional human review for real money / sensitive PII at scale.
Anti-Patterns
- Printing secret values in the report.
- Claiming the app is "secure."
- Inflating potential findings into confirmed.
- Skipping a check "to save time/tokens."
Validation Checklist
Definition of Done
A security review report with severity-rated, Confirmed/Potential findings (secrets redacted), a rotation list, verification notes, and — for money/PII-at-scale — a human-review recommendation.
Related Skills
code-review, dependency-audit, environment-audit, ai-output-review, final-quality-audit, release-planning, project-orchestrator, project-management.
Related Knowledge
../../knowledge/ (data model, trust boundaries).
Related References
../../references/security/ (if present) for baselines.
Context Loading Guidance
- Requires: the security-relevant surface, data-sensitivity context.
- Does not require: unrelated modules, the full reference tree, planning skills.
- May load:
dependency-audit, environment-audit for exposure.
- Stop when: the report (with rotation list) is delivered.
Token Efficiency Guidance
Focus on entry points, auth, data access, and config. Quote decisive lines, redacted; don't paste whole files or any secret value.
1---2name: security-review3description: Use to assess a change or codebase for security weaknesses — secrets, auth, authorization/IDOR, injection, PII handling, transport/errors, payments, and production hardening. Reports findings by severity, separates confirmed from potential, never prints secret values, and never claims "secure."4metadata:5 source: https://skillmd.com/skills/ahtishamshahzad/security-review6---78# Security Review910## Purpose1112Find real, exploitable security weaknesses and report them safely. Implements `../../system/SECURITY_RULES.md`. Supports Gate 6 and any security-audit request.1314## When to Use1516- Before merge/release for changes touching auth, data, payments, or external input.17- For a **security audit** request type.18- Whenever an audit or review surfaces a security-relevant surface.1920## Inputs2122- The change/codebase in scope and how it handles secrets, auth, data.23- Whether the app handles money or sensitive PII.2425## Discovery Questions2627- Does the change touch secrets, auth, authorization, input handling, PII, or payments?28- Is this pre-launch, live, or post-incident?29- What is the trust boundary and the data sensitivity?3031## Responsibilities3233Assess, per `../../system/SECURITY_RULES.md`:34- **Secrets** (no hardcoding; git history; client-exposed prefixes; correct key placement).35- **Auth** (enforced routes, JWT/session integrity, reset flow, rate limits, no default admin).36- **Authorization/IDOR** (ownership checks, server-side roles, tenant scoping).37- **Input/injection** (parameterized queries, XSS, command, uploads, validation).38- **PII** (minimized in logs/third-party; deletion path).39- **Transport/errors** (TLS, CORS, security headers, generic errors).40- **Payments** (server-side totals, webhook signatures, abuse) when applicable.41- **Production hardening** (debug/test endpoints, env fail-fast, dependency advisories).4243Report findings with **severity** and **Confirmed vs Potential**.4445## Required Workflow46471. Scope to the change/surface + data sensitivity.482. Walk the relevant security areas. Check that security behavior is specified as scenarios — cross-tenant and cross-user denial, expired or invalid tokens, role escalation refused (`../../system/GHERKIN_RULES.md`) — and that each has a passing test. A scenario states the expectation; only tests and review are evidence. Never conclude "secure".493. Run available tools where possible; quote or mark "unverified until run."504. Record findings (severity, Confirmed/Potential, `file:line`, redacted).515. List secrets to rotate; recommend human review for money/PII at scale.526. On a tracked project, record them via `project-management`: each Confirmed finding through bug intake (Critical/High → P0/P1), Potential ones as `TECH-` tasks to verify (`../../system/PROJECT_MANAGEMENT_RULES.md`, recording work).5354## Decision Rules5556- Never print a discovered secret's value — report location/type, redacted, and flag for rotation.57- Separate Confirmed from Potential; assign severity (Critical/High/Medium/Low).58- Block release on open Critical/High confirmed issues.59- Never claim "secure"; report what was checked and fixed, and what needs human review.6061## Rules6263- Do no harm: no data exfiltration, no hitting production/third-party systems, no destructive commands.64- Do not weaken existing controls to "make it work."65- Recommend a professional human review for real money / sensitive PII at scale.6667## Anti-Patterns6869- Printing secret values in the report.70- Claiming the app is "secure."71- Inflating potential findings into confirmed.72- Skipping a check "to save time/tokens."7374## Validation Checklist7576- [ ] Relevant security areas assessed.77- [ ] Findings: severity + Confirmed/Potential + redacted location.78- [ ] Secrets-to-rotate list produced.79- [ ] Tools run or unrun checks flagged.80- [ ] Human-review recommendation where warranted.81- [ ] No secret values printed; no "secure" claim.8283## Definition of Done8485A security review report with severity-rated, Confirmed/Potential findings (secrets redacted), a rotation list, verification notes, and — for money/PII-at-scale — a human-review recommendation.8687## Related Skills8889`code-review`, `dependency-audit`, `environment-audit`, `ai-output-review`, `final-quality-audit`, `release-planning`, `project-orchestrator`, `project-management`.9091## Related Knowledge9293`../../knowledge/` (data model, trust boundaries).9495## Related References9697`../../references/security/` (if present) for baselines.9899## Context Loading Guidance100101- **Requires:** the security-relevant surface, data-sensitivity context.102- **Does not require:** unrelated modules, the full reference tree, planning skills.103- **May load:** `dependency-audit`, `environment-audit` for exposure.104- **Stop when:** the report (with rotation list) is delivered.105106## Token Efficiency Guidance107108Focus on entry points, auth, data access, and config. Quote decisive lines, redacted; don't paste whole files or any secret value.