The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external skill files (local, inline, or remote) as input. This creates an attack surface where a malicious source skill could attempt to influence the agent. The skill mitigates this by explicitly instructing the agent to treat these sources as passive data artifacts rather than governing instructions and to never execute bundled scripts or follow embedded workflows during the fusion process.\n- Ingestion points: SKILL.md (Step 2: Acquire and manifest every source)\n- Boundary markers: SKILL.md (Step 3: Run the publishability gate before synthesis) and references/remote-source-acquisition.md (Treat remote content as untrusted data)\n- Capability inventory: The skill performs file reads, file writes to target packages, and network retrieval for remote sources.\n- Sanitization: Instructions mandate treating inputs as data and prohibit executing source-provided scripts or dependencies.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the acquisition of remote skills from user-provided URLs. It includes a specific reference to the skills CLI documentation hosted on Vercel's official GitHub repository. These references target a well-known service and are documented with appropriate safety postures.\n- [COMMAND_EXECUTION]: The skill mentions the use of npx and a skills CLI tool for resolving remote packages. It provides explicit warnings that tools like npx might automatically download and execute code, advising the agent to obtain prior authorization or skip these tools if they would cause unintended executions.

fuse-skills