The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute shell commands for environment discovery and project management.
Uses find starting from the user's $HOME directory to locate git repositories across common development paths.
Executes git commands (log, show, status, push, checkout) to interact with project history and remote servers.
Uses the GitHub CLI (gh pr create) to automate pull request creation.
[CREDENTIALS_UNSAFE]: The skill attempts to read the user's shell configuration file (~/.zshrc) to retrieve environment variables such as FH_DIR and CC_HUB_DIR. Shell profiles are sensitive files that frequently contain plain-text credentials, API keys, and private configuration.
[DATA_EXFILTRATION]: The skill collects data from local repositories, including commit history, code diffs, and session summaries, and transmits this data to external repositories.
Uses git push to upload harvested patterns to remote branches.
Uses gh pr create to send session data and code patterns to the GitHub platform.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from git commit messages and code diffs, creating a surface for indirect prompt injection.
Ingestion points: Git commit logs and file diffs are read via git log and git show commands as defined in SKILL_detail.md (§ModeA-Blocks and §ModeB-Blocks).
Boundary markers: Absent; the skill directly aggregates commit messages into PR bodies and session logs without delimiters or "ignore embedded instructions" warnings.
Capability inventory: Significant capabilities including Bash (filesystem access, git operations, GitHub CLI), Write, and Edit permissions across the local environment.
Sanitization: No sanitization, escaping, or validation of the commit content is performed before interpolation into pull requests or hub session files.

field-harvest