Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
DeepSeek
Muse+22 moreClaude CodeCursorCodex CLIDeepSeekMuse
Dots
Manus
Grok
Grok Bot+18 agentsSecurity Secrets & Dependency Audit — Leak Risk, Attack Path
The problem Scanner dumps list “possible secrets” without saying what the business loses or what an attacker would do next.
$15
Security Secrets & Dependency Audit — Leak Risk, Attack Path
Example session with this skill installed
Mode: Gate (publish Apache-2.0 next week)
Repo: glacier-archive-worker — Go media archive indexer (university)
Surfaces
- docker-compose.override.yml with R2-style access key pair + prod bucket name (~5 months in git)
- helm/glacier/values-prod.yaml with smtps://apikey:SG.…@smtp… embedded
- worker.example.toml YOUR_R2_ACCESS_KEY_HERE (placeholder)
- docs/setup.md glpat-ExamplePlaceholderToken0000 (placeholder)
- go.mod: gopkg.in/yaml.v2 v2.2.8, golang.org/x/net v0.7.0
Ask: go/no-go for open-source; mask all secrets; do not modify the repo.
- Read your context and instructions
- Compiled the security secrets
Security Secrets & Dependency Audit
1. Assessment Summary
| Field | Value |
|---|---|
| Repo | glacier-archive-worker |
| Mode | Gate (pre-open-source) |
| Threat model | Object-store + SMTP + researcher emails — high if keys real |
| Secrets | 4 (2 critical patterns / 2 placeholders) |
| Dependencies | 2 (verify via govulncheck) |
| Maturity | CRITICAL-EXPOSURE |
| Open-source gate | NO-GO |
2–3. Threat model & surfaces
Research media store + institutional email identity. Compose, Helm, docs, go.mod scanned.
Placeholders filtered; compose + Helm fail Gate.
4. Secrets (masked only)
| ID | Where | Masked | Class | Severity |
|---|---|---|---|---|
| SEC-01 | docker-compose.override.yml | AKIA**** / wJal**** | needs-review → treat compromised | CRI |
| SEC-02 | helm values-prod | SG.**** in SMTP URL | real-pattern | CRITICAL |
| SEC-03 | example.toml | YOUR_R2_**** | placeholder | INFO |
| SEC-04 | docs/setup.md | glpat-**** | placeholder | INFO |
Lifecycle: SEC-01 ~5 months in remote history — rotation required even if file is deleted.
5. Dependencies
DEP-01 yaml.v2@v2.2.8 — outdated; verify advisories / upgrade
DEP-02 x/net@v0.7.0 — outdated; govulncheck before release
(No invented CVE IDs.)
6–7. Attack paths & business impact
- Open-source × SEC-01 → keys tried against research bucket
- SEC-02 → send-as phishing
- SEC-01 + reachable parser issue (if confirmed) → chained host impact
Priority: SEC-01, SEC-02, then deps.
8–9. First hour + roadmap
Rotate object-store + SMTP keys now → audit provider logs → assume history leaked →
externalize Helm secrets → govulncheck upgrades → pre-commit/CI gates → re-run Gate.
10–13. Prevention, confidence, maturity
Layers: pre-commit → CI → secrets manager → no tracked overrides → Renovate/govulncheck.
Verdict CRITICAL-EXPOSURE until critical secrets gone; then HIGH-RISK → MODERATE → CLEAN path.
Static only — not a pen test.
14. JSON
Full Audit JSON with masked_value fields, attack_paths, remediation_roadmap, assumptions.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
- Scanner dumps list “possible secrets” without saying what the business loses or what an attacker would do next.
- Credentials in Docker Compose overrides, Helm values, and CI history survive long after a file is “fixed.”
- Dependency noise (dev-only, unreachable) drowns out reachable, high-impact issues.
- Teams open-source or share repos without a clear go/no-go gate.
What it does
- Scopes a threat model (what the system protects and why findings matter).
- Scans major leak surfaces: source, config, .env, IaC, CI/CD, Docker, Helm/k8s, docs, fixtures.
- Classifies secrets as real / placeholder / needs-review and masks values (e.g.
AKIA****MPLE). - Adds secret lifecycle: age, git history, remote exposure, environment.
- Weights dependencies by severity, reachability, exploitability, maintainer health, and license.
- Builds attack paths and a business-impact order.
- Delivers first-hour response, remediation roadmap, prevention layers, and a maturity verdict with a path up.
Frameworks & tools
- Works across Node, Python, Go, containers, Helm, and common package manifests (
package.json,go.mod,requirements.txt, etc.). - Modes: Audit, Incident, Gate (pre-release/open-source), Compare (baseline).
- Compatible with Claude Code, Cursor, Codex CLI, OpenClaw — file-read only for the assessment itself.
Why this beats prompting it yourself
- False-positive discipline so
YOUR_KEY_HEREdoes not page the same as a live key. - Lifecycle thinking: history and remotes matter more than “file deleted.”
- Attack paths, not flat lists.
- Explicit open-source / release Gate verdict.
- Safety-by-design: no full secret dump in the report.
- Honest coverage map when advisories were not machine-verified.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
• Static analysis only — not a pen test or runtime proof
• Without git history, lifecycle analysis is limited
• Without lockfiles / govulncheck, dependency confidence is lower
• Does not invent CVE IDs; uncertain items are marked for advisory verification
• Does not rotate credentials or rewrite git history for you
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 21 days ago
Permissions
This skill only reads project files you provide. It drafts a risk report with masked secret fragments. It does not rotate keys for you, call cloud APIs, or upload your code. Grant Read only.