The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the gh (GitHub CLI) tool via subprocess.run to interact with GitHub APIs and fetch repository metadata. The implementation consistently uses list-based arguments without shell=True, which is a recommended practice to avoid command injection vulnerabilities.
[DATA_EXFILTRATION]: The skill collects and aggregates user profile data, including names, companies, and email addresses (public and commit-derived), from GitHub repositories. This information is stored in local CSV files for lead generation purposes. While this involves handling sensitive contact information, it is the primary stated functionality of the skill and is performed under user direction using their own authentication.
[PROMPT_INJECTION]: The skill ingests untrusted data from GitHub repositories, such as user bios, issue titles, and comments, which are then analyzed by the agent. This creates a potential surface for indirect prompt injection where malicious content in a profile or issue could attempt to influence the agent's behavior.
Ingestion points: Untrusted strings are fetched from GitHub APIs in scripts/gh_repo_signals.py, scripts/gh_contributors.py, and scripts/gh_issues_scanner.py.
Boundary markers: None identified. Data is aggregated into structured CSVs, but the agent is not explicitly instructed to treat the content as untrusted during its analysis phase.
Capability inventory: The skill allows the agent to execute shell commands (via gh) and perform file system operations.
Sanitization: No filtering or escaping is applied to user-provided strings from GitHub before they are presented to the agent for review.

github-repo-signals