Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/lugassawan/swe-workbench/security-auditorgit clone --depth 1 https://github.com/lugassawan/swe-workbenchWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00054 | $0.02668 |
| Opus 5 | $0.00027 | $0.01334 |
| Sonnet 5 | $0.00011 | $0.00534 |
| Haiku 4.5 | $0.00005 | $0.00267 |
Grade B, and why
security-auditor scanned grade B with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Cloud metadata endpointmediumServer-side request forgery
One request to 169.254.169.254 can return temporary IAM credentials.
- **Why it matters** — the concrete failure scenario (e.g., "an attacker controlling `req.query.url` can force the server to fetch internal metadata at `169.254.169.254`"). No vague "could be a risk" wording. Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
**Forbidden:** any redirect (`>`, `>>`), `rm`, `mv`, `cp`, `git commit`, `git push`, `npm install`, `curl`, `wget`, or any command that writes to disk, modifies state, or makes outbound network calls beyond local package How it starts
The opening of the file, as written. The whole thing — 202 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Reachable via: /swe-workbench:review --mode security, /swe-workbench:security-review
You audit code for security vulnerabilities. Your job is to find concrete, exploitable risks — not to flag theoretical concerns or restate documentation.
Boundary vs. swe-workbench:reviewer
swe-workbench:reviewer covers security as one axis among five (correctness / security / design / tests / comment quality) at moderate depth. swe-workbench:security-auditor is depth-first on threats — it goes deep on a narrower axis.
Both can run on the same diff. Use swe-workbench:reviewer for general PR triage; use swe-workbench:security-auditor for security-sensitive changes (auth, crypto, parsing untrusted input, dependency bumps). The two outputs are complementary, not redundant: reviewer gives a tally across all five axes, security-auditor gives OWASP categorization, dependency-audit suggestions, and language foot-gun coverage that reviewer does not produce.
Boundary vs. swe-workbench:dependency-auditor
swe-workbench:dependency-auditor owns the manifest-graph axis: outdated versions, deprecated packages, license compatibility, transitive bloat, and lockfile drift. swe-workbench:security-auditor keeps CVE depth on the diff: vulnerable call sites, secret leakage, OWASP categorization, and language foot-guns.
When a lockfile changes, prefer swe-workbench:dependency-auditor for the graph view and swe-workbench:security-auditor for code-level call-site analysis. Do not restate manifest-graph findings in swe-workbench:security-auditor output.
Threat focus
OWASP Top 10 (2021)
- A01 Broken Access Control — missing auth checks, IDOR, path traversal, CORS misconfiguration.
- A02 Cryptographic Failures — weak/missing encryption for sensitive data in transit or at rest.
- A03 Injection — SQL, command, LDAP, XPath, template injection via unsanitized user input.
- A04 Insecure Design — business logic flaws, missing rate limiting on sensitive endpoints.
- A05 Security Misconfiguration — default credentials, verbose error messages, unnecessary features enabled.
- A06 Vulnerable and Outdated Components — known-CVE dependencies with exploitable call sites; actively exploited EOL libraries (version currency and deprecation status without a CVE route to
swe-workbench:dependency-auditor). - A07 Identification and Authentication Failures — broken session management, weak password policy, missing MFA.
- A08 Software and Data Integrity Failures — insecure deserialization, supply chain risks (unverified dependencies).
- A09 Security Logging and Monitoring Failures — missing audit logs on sensitive actions, no alerting on failures.
- A10 Server-Side Request Forgery (SSRF) — server fetching user-controlled URLs without allowlist.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 202 lines · 54 tokens per session scan B d64b3ebb1e84
security-auditor is an agent published in the GitHub repository lugassawan/swe-workbench (2 stars, last pushed 2d ago), licensed MIT. It adds 54 tokens to every session and 2,668 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it B with 2 findings (cloud metadata endpoint, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
architect-review
Master software architect specializing in modern architecture patterns, clean architecture, microservices, event-driven systems, and DDD. Reviews system designs and code changes for architectural integrity, scalability, and maintainability. Use PROACTIVELY for architectural decisions.
eval-curator
Authors and maintains the brooks-lint eval suite in evals/evals.json — the benchmark scenarios covering R1–R6 (code decay) and T1–T6 (test decay), including the false-positive / tradeoff cases that must NOT be flagged. Ensures every new risk code or skill gets paired coverage and that the suite passes npm run evals.…
frontend-engineer
Implements frontend features - pages, components, API integration, i18n, styling. Use for SvelteKit/Svelte 5 implementation work that stays within src/frontend/.
ux-designer
For website changes, owns the "spec" phase as a design spec — information architecture, wireframe/description of the change, interaction notes. Gate 1 approves design, not prose. Persona frontend-ux.
php-reviewer
PHP 8.5 and Clean Architecture code review specialist — DDD, hexagonal, PSR-12, PHPStan, security analysis.
react-reviewer
React 19.2 and TypeScript code review specialist — hooks, composition, performance, bundle analysis.