The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from a remote catalog and user-supplied descriptions to generate agent instructions, creating a surface for indirect prompt injection.
Ingestion points: Data enters the agent context from the bundled references/catalog.md, a remote live catalog at signals.forwardfuture.ai, and user-provided input during design interviews.
Boundary markers: The instructions explicitly command the agent to treat live content as untrusted reference data that cannot override the skill's core instructions, repository policies, or user constraints.
Capability inventory: The skill is designed to output text-based prompts and does not possess capabilities for arbitrary code execution, file system modification, or network exfiltration of sensitive data.
Sanitization: The skill directs the agent to summarize and adapt catalog content under strict guardrails rather than copying or executing remote instructions verbatim.
[EXTERNAL_DOWNLOADS]: The skill is configured to fetch an updated catalog from an external URL upon explicit user request.
The target domain signals.forwardfuture.ai is owned by the author (Forward Future) and serves as a repository for reference data.
The download is restricted to data retrieval (JSON/Markdown) and does not involve the execution of remote scripts.

loop-library