The Agent Skills Directory

[SAFE]: The detect-stack.sh script is a POSIX-compliant shell script that performs local file system analysis using standard tools (find, grep, sed). It explicitly avoids network operations and limits its scope to identifying common programming ecosystems (Node.js, Python, Go, Rust, Java). The script includes a PRUNE list to prevent traversing sensitive or redundant directories like .git or node_modules. No command injection vulnerabilities were found, as the script handles file paths and arguments carefully using standard shell quoting practices.
[SAFE]: The SKILL.md file defines a structured workflow for plan analysis, including mandatory coverage checks and evidence-based safeguard triggers. It emphasizes a 'review-only' mode when input sources are not safely editable and requires human-in-the-loop triage before any modifications are proposed. The inclusion of a research floor section requires agents to cite normative documentation, reducing the risk of hallucinated or malicious advice.
[SAFE]: The provided fixtures serve as regression tests for the agent's logic, ensuring it correctly identifies issues like missing authentication (safeguard-gap) or speculative abstractions (absent-consumer). These files do not contain malicious payloads and are used to validate the consistency and safety of the agent's output.

leanplan