The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from directory listings, scraped web content, and third-party exports. It includes specific defensive instructions to treat this information as untrusted and to ignore any embedded instructions found within the data.\n
Ingestion points: Web scrapes, directory content, emails, and third-party exports as identified in the 'Data Sources' and 'Instructions' sections of SKILL.md.\n
Boundary markers: Present. Explicit instructions in the 'Data Sources' and 'Skill Contract' sections warn the agent to treat scraped material as evidence and not as instructions.\n
Capability inventory: No executable capabilities or scripts were found in the provided files.\n
Sanitization: Present. The skill instructs the agent to verify destination legitimacy, check for lookalike domains, and avoid installing helper tools or supplying credentials to listing pages.\n- [NO_CODE]: The skill consists entirely of instructional Markdown files and configuration. No executable scripts (Python, JavaScript, Shell) or binaries are included, which significantly reduces the attack surface for most technical threats.\n- [CREDENTIALS_UNSAFE]: No hardcoded API keys or secrets were found. The skill documentation explicitly advises keeping credentials out of source control and using dedicated environment variables or secret management tools.\n- [SAFE]: The skill adheres to the principle of least privilege by requiring explicit user authorization for any action that could result in persistent changes, account mutations, or external side effects.

directory-submissions