The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository data (commits, code, and filenames) which creates a surface for indirect prompt injection. Malicious content within the analyzed repository could attempt to influence the agent's audit verdict.
Ingestion points: The agent reads repository metadata, commit logs, and file content across all branches and tags as specified in the Preflight and Pass sections of SKILL.md and references/full-guide.md.
Boundary markers: No specific delimiters or boundary markers are defined to isolate untrusted repository data from the agent's instructional context.
Capability inventory: The skill utilizes shell command execution (via git, grep, and various auditing tools) to process the repository content.
Sanitization: Content retrieved from the repository is not sanitized or filtered before being processed by the agent.
[COMMAND_EXECUTION]: The skill executes multiple shell commands to perform its audit functions. This includes standard git operations (git fetch, git log, git stash) and file system searches using grep and find. It also invokes specialized security CLI tools such as gitleaks and trufflehog for secret detection.
[EXTERNAL_DOWNLOADS]: The skill utilizes external auditing tools via package managers. It invokes bunx license-checker for JavaScript dependency analysis and references pip-licenses, cargo-license, and go-licenses for other ecosystems. These tools are fetched from well-known, established package registries.

open-source-checker