The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The setup script programmatically scrapes authentication secrets from local user configuration files.
Evidence: Extracts GITHUB_TOKEN from ~/.robin/.env using grep and cut.
Evidence: Parses ~/.git-credentials to extract plain-text credentials for github.com.
[COMMAND_EXECUTION]: Employs inline Python scripts to process data retrieved from the network.
Evidence: Multiple uses of python3 -c to parse JSON output from curl commands.
[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by ingesting and displaying untrusted content from GitHub.
Ingestion points: The skill fetches issue titles, bodies, and comments via the GitHub API in SKILL.md.
Boundary markers: None. Content from external issues is printed directly into the agent's context without delimiters or warnings.
Capability inventory: The skill can perform network operations (curl), filesystem operations (git), and execute arbitrary logic via python3.
Sanitization: None. Data from fields such as i['body'] and i['title'] are printed verbatim.

github-issues