The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from GitHub issue descriptions and comments, which could potentially contain malicious instructions intended to manipulate the agent's behavior.
Ingestion points: The skill retrieves issue metadata, bodies, and comments using the gh issue view command (SKILL.md, steps 3-5).
Capability inventory: The skill has the ability to execute shell commands (git, npm, gh), perform file system modifications, and interact with remote repositories (SKILL.md, steps 2, 7).
Boundary markers: The instructions do not specify any delimiters or safety warnings to ignore instructions found within the issue data.
Sanitization: There is no explicit logic described for sanitizing or filtering external content before processing.
Mitigation: A critical safety feature is the mandatory user approval step (Step 6) where the agent must present its plan and wait for an 'OK' before executing any implementation, commits, or issue closures.
[COMMAND_EXECUTION]: The skill executes various local system commands to manage the software development lifecycle.
Evidence: Uses git for branch management and pushing changes, gh for interacting with GitHub issues, and node/npm for versioning and installing dependencies (SKILL.md, steps 1, 2, 3, 7).
[EXTERNAL_DOWNLOADS]: The skill initiates downloads of external packages during its setup phase.
Evidence: The skill runs npm install (SKILL.md, step 2), which fetches dependencies from the public NPM registry. This is standard for Node.js project maintenance.

resolve-issues-cx