The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs git push operations which transmit local repository content to a remote server. To mitigate data exposure, the instructions explicitly command the agent to unstage sensitive files such as .env*, credential files, and local secrets before committing.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from multiple sources to generate pull request content.
Ingestion points: Reads git status, git diff, and git log, and fetches issue titles and URLs from external trackers (SKILL.md).
Boundary markers: Uses a structured Markdown template for the PR body to separate different types of information.
Capability inventory: Executes shell commands (git, gh) and potentially project-specific installation/test scripts.
Sanitization: Implements an "out-of-scope" list check for diffs and explicitly excludes credentials/secrets from the staging process.
[COMMAND_EXECUTION]: The skill utilizes several git commands (status, diff, log, add, push) and CLI tools for pull request management. It also executes "configured" scripts for environment installation and local validation gates.
[DYNAMIC_EXECUTION]: The skill dynamically identifies and executes project-specific commands for installation (run the configured install) and testing (run the configured full local gate) based on the repository's local configuration files.

ziw-pr