The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data retrieved from Lore transcripts.
Ingestion points: Transcripts are ingested via lore search, lore get, lore context, and lore session CLI commands.
Boundary markers: The skill does not define specific input delimiters but requires a highly structured JSON output as defined in references/evidence-packet.md to prevent schema confusion.
Capability inventory: Subprocess calls are limited to the lore CLI and a local Node.js validation script.
Sanitization: The instructions explicitly state that transcripts are "testimony, not truth" and require current system state to override stale transcript claims. It also mandates paraphrasing when transcript content is sensitive.
[COMMAND_EXECUTION]: The skill utilizes local CLI commands (lore) and Node.js to perform its core functions and validation. These commands are part of the intended workflow and do not involve executing external or unverified scripts.

lore-recall