The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates workflows involving the processing of untrusted external data such as GitHub issue descriptions and pull request content. This represents a potential surface for indirect prompt injection where malicious instructions hidden in these data sources could attempt to influence the agent.
Ingestion points: The skill is designed to handle content from GitHub issues and pull requests as described in references/setup.md and references/pull-request.md.
Boundary markers: The instructions do not define specific delimiters or "ignore" instructions to isolate external data from the agent's instruction context.
Capability inventory: The skill utilizes system-level tools including bash for verification scripts, git for repository management, and gh (GitHub CLI) for issue and PR creation.
Sanitization: There are no explicit sanitization or filtering steps mentioned for external data before it is incorporated into the agent's context.
[COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands and scripts local to the repository for environment setup and verification.
Evidence: Instructions include running bash scripts/verify.sh, ./scripts/pre-commit-setup.sh, and uv pip install -e ".[dev]". These are standard developer operations within the context of the repository.

quantmind-dev