The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands to interact with the GitHub API and manage local cache files.
Utilizes the gh CLI for authenticated API calls to GitHub's REST and GraphQL endpoints.
Uses jq for parsing, filtering, and structuring JSON responses.
The scripts/github-activity.sh script performs file system operations including mkdir, mv, and rm to manage a per-day cache in ~/.cache/recap/github/.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub that could be manipulated by third parties to influence agent behavior.
Ingestion points: Fetches user-provided strings such as PR titles and issue descriptions via gh api search/issues and GraphQL queries in scripts/github-activity.sh.
Boundary markers: The skill stores the fetched data in standard JSON format without encompassing the content in delimiters or including instructions for the agent to ignore embedded commands.
Capability inventory: The skill provides access to network resources (GitHub API) and local command execution capabilities (gh, jq, date).
Sanitization: There is no evidence of sanitization or filtering of the fetched strings before they are stored in the local cache, allowing potential malicious instructions in PR/Issue content to be persisted.

github-activity