The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts to manage its state and environment. SKILL.md contains multiple shell blocks that invoke the bundled sidecar.sh script to resolve file paths, manage work items, and perform repository lookups.
[EXTERNAL_DOWNLOADS]: The skill interacts with remote GitHub repositories using the official CLI. Instructions include using gh api and gh issue list to retrieve metadata, project milestones, and technical research artifacts from GitHub.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository and user input, creating a surface for potential injection attacks.
Ingestion points: The skill reads project research documents, past solutions from the filesystem, and explores the codebase to understand the current technical state.
Boundary markers: The instructions lack explicit safety delimiters or 'ignore' instructions for the interpolated repository content, which could allow malicious instructions in processed files to influence agent behavior.
Capability inventory: The associated sidecar.sh script can write files to the local filesystem and perform git commits, while the GitHub CLI integration allows for creating and editing issues.
Sanitization: The skill employs jq for structured JSON data handling and a sanitization routine for branch names, which mitigates simple command injection risks during data processing.

write-a-prd