The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a suite of local bash scripts (e.g., detect-stack.sh, analyze-complexity.sh) to perform static analysis on a target project. These scripts use standard Unix tools like find, grep, awk, and sed to identify patterns and metrics.
[PROMPT_INJECTION]: The skill aggregates codebase content—including TODO comments, documentation, and source code—into an audit report analyzed by the LLM, creating a potential surface for Indirect Prompt Injection.
Ingestion points: File content and metadata are read from the audited project by scripts in the assets/ directory (e.g., assess-tech-debt.sh, detect-security-risks.sh).
Boundary markers: generate-audit-report.sh encapsulates script outputs within Markdown code blocks and YAML delimiters to provide structural separation for the model.
Capability inventory: The skill possesses local file read and shell execution capabilities required for its auditing functions; it does not have network access or file-write capabilities in the audit pipeline.
Sanitization: Ingested content from the audited codebase is not sanitized or filtered for instructions before being passed to the LLM.
[SAFE]: The skill follows security best practices by explicitly excluding sensitive directories (such as .git, node_modules, and vendor) from its scans and avoiding the use of external dependencies or remote downloads. The detection of hardcoded secrets is a core intended feature and is conducted entirely locally.

skill-security-checklist