The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local shell script at scripts/recap.sh which in turn executes a binary named ch-recap using a platform-provided resolution script.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it ingests and processes untrusted data from historical conversation logs.
Ingestion points: Historical conversation transcripts are retrieved via the recap.sh command and then read by the agent to generate summaries (SKILL.md).
Boundary markers: The transcript output format uses specific delimiters like === Conversation ... === and timestamp headers to structure the data, though these do not prevent the agent from following embedded instructions.
Capability inventory: The skill is intended for summarization and references another skill for loading detailed conversation logs based on the recap findings.
Sanitization: There is no evidence that the content of the historical transcripts is sanitized, filtered, or escaped before being presented to the agent for processing.

conversation-recap