The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of data from the InterPlanetary File System (IPFS) via the retrieve action. Since IPFS is a decentralized network where any user can host content, data retrieved by a Content Identifier (CID) is untrusted and could contain adversarial instructions designed to influence the agent.
Ingestion points: Untrusted content enters the agent's context through the retrieve action documented in SKILL.md and schema.md.
Boundary markers: The skill instructions do not specify the use of delimiters or clear "ignore embedded instructions" warnings for data returned from IPFS.
Capability inventory: The skill possesses network read (retrieve, list) and network write (upload) capabilities through the AgentPMT-hosted tool.
Sanitization: No explicit sanitization, filtering, or validation of the retrieved content is mentioned in the skill documentation.
[EXTERNAL_DOWNLOADS]: The installation section provides instructions to fetch skill components from the vendor's GitHub repository (AgentPMT/agent-skills) using platform-standard installers.
[COMMAND_EXECUTION]: The documentation includes shell commands for the npx and openclaw CLI tools to facilitate the installation and management of the skill.
[NO_CODE]: The skill is composed of markdown instructions and JSON schemas for remote tool calls; it does not distribute local executable scripts or binaries.

ipfs-content-gateway