Works with the AI Agents you already use
Security Workbench: code security audit with recorded decisi
Structured security auditing and verified cryptography implementation from research papers.
$30
Security Workbench: code security audit with recorded decisi
Example session with this skill installed
Run a security scan on this repository to find hardcoded secrets and vulnerable dependencies. I've already installed the requirements.
- Read your context and instructions
- Compiled the security workbench
I have initiated the sec-scan stage to assess the repository. The scanner evaluated the local filesystem, checked dependency versions against the OSV database, and searched for high-entropy strings.
Security Scan Findings
| ID | Finding | Severity | Evidence Level | Location |
|---|---|---|---|---|
| SEC-001 | Hardcoded AWS Access Key | Critical | pattern-match | src/config/aws_provider.py:12 |
| SEC-002 | Unpinned Alpine Base Image | Low | tool-verified | docker/Dockerfile:1 |
| SEC-003 | Use of os.system with user input | High | model-judgment | src/utils/shell.py:45 |
| SEC-004 | requests v2.25.0 (CVE-2021-33503) | Medium | tool-verified | requirements.txt:4 |
Policy Decisions Required
The scanner found one critical secret pattern. Per team policy, please confirm if SEC-001 is a valid credential or a false positive.
Next steps
- Review the AWS key in
src/config/aws_provider.pyand rotate it immediately if valid. - Update
requirements.txttorequests>=2.31.0to resolve the known vulnerability. - Run
sec-auditif you require a formal report including a threat model and full inventory.
Connects securely to your tools. The creator never sees your data.
About this skill
Manual security reviews are slow and prone to oversight, especially when implementing complex cryptographic standards. This skill provides a structured framework for code auditing and formal cryptography implementation, moving from paper specs to verified code.
What it does
- Code scanning identifies hardcoded secrets, injection points, and unsafe API patterns using tiered evidence levels.
- Dependency auditing checks unpinned container images and queries vulnerability databases for known risks.
- Cryptography implementation translates research papers, RFCs, or standards into specifications and tested code.
- Audit reporting generates structured findings with priorities, requiring human verification for every issue.
- Evidence tracking categorizes findings as tool-verified, test-verified, pattern-matched, or model-judgments.
How it works
- Initialize the workbench in a project directory to create the local configuration and state tracking.
- Execute scan for a rapid assessment of high-priority security flaws and leaked credentials.
- Run audit for a deep-dive review involving threat modeling, checklists, and dependency verification.
- Implement crypto by following the pipeline from intake and spec extraction to secure coding and final review.
Frameworks & tools
The workbench utilizes Python 3.11+ and PyYAML. It interacts with external advisory databases like OSV (with consent) and requires standard developer tools for running project-specific test vectors and fuzzers.
Why this beats prompting it yourself
Generic prompts often miss unpinned dependencies or provide confident but incorrect cryptographic implementations. This skill enforces a multi-stage verification pipeline that treats findings as candidates until verified by tools or tests, preventing silent assumptions.
Use cases
- Scan a legacy repository for hardcoded API keys and unpinned Docker images.
- Implement a new cryptographic primitive from a formal academic paper or RFC.
- Conduct a pre-release security audit including dependency checks and threat modeling.
- Compare multiple research papers to synthesize a single secure design for a project.
Known limitations
Requires Python 3.11 and PyYAML for scanner modules. Network access to vulnerability databases and execution of project code for testing require explicit user consent for each session.
Requires Python 3.11+ and PyYAML. Network lookups and running project code for tests happen only with your approval. Secrets the rules recognize are masked in all output (console, SARIF, JSON), so results can be stored and shared
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 any payment.
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 24 days ago
Requires
Permissions
Allowed Hosts