The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which could contain malicious instructions designed to override agent behavior.
Ingestion points: Governed project-agent chat, project library records, and TreeDX projections (references/knowledge.md, references/execution.md).
Boundary markers: Present. Instructions mandate the use of input_required signed confirmation checkpoints and strictly following server-provided structured schemas (SKILL.md).
Capability inventory: The skill uses the trsd CLI for querying, searching, and managing project libraries and workspaces (references/knowledge.md).
Sanitization: Instructions explicitly forbid placing credentials, private keys, or bearer tokens in tool arguments or model-visible content (SKILL.md, references/mcp.md).
[EXTERNAL_DOWNLOADS]: The skill references and fetches verification manifests from vendor-owned infrastructure.
Evidence: scripts/verify.py uses urllib.request.urlopen to download contract-bundle.json and component-release.json from the vendor's GitHub repository (treeseed-ai) to verify integrity hashes.
[COMMAND_EXECUTION]: The skill provides instructions for utilizing the vendor's trsd command-line utility for library management.
Evidence: references/knowledge.md documents various trsd library subcommands for searching, reading, and organizing project knowledge and workspaces.

treeseed