The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/fetch_skills.py contains a hardcoded API key (sk_live_skillsmp_...) for the SkillsMP marketplace. Hardcoding credentials in shared code is a security risk, even if intended for public access.\n- [EXTERNAL_DOWNLOADS]: The skill fetches data from the GitHub API and skillsmp.com. This involves communicating with external community repositories to provide search results and deep-dive details.\n- [COMMAND_EXECUTION]: The skill uses subprocess.run to call the GitHub CLI (gh). This is used to perform authenticated API queries when the tool is installed in the user environment.\n- [PROMPT_INJECTION]: The skill exposes the agent to indirect prompt injection by retrieving and displaying the content of remote SKILL.md files from external repositories.\n
Ingestion points: The deep_dive function in scripts/fetch_skills.py retrieves file content from remote GitHub repositories.\n
Boundary markers: The skill uses basic separators like --- SKILL.md ---, but provides no explicit safety instructions to the agent regarding the remote content.\n
Capability inventory: The agent context has capabilities for command execution and file manipulation.\n
Sanitization: Remote content is displayed raw without sanitization or instruction-filtering.

skill-installer