The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of potentially attacker-controlled data into the agent context through the evidence gathering process.
Ingestion points: Step 1 in SKILL.md specifies reading alerts, monitoring metrics, chat transcripts, and deploy history from the host system.
Boundary markers: The instructions do not mandate the use of delimiters or explicit warnings to ignore embedded instructions within the gathered evidence.
Capability inventory: The skill leverages write, edit, and exec tools to create and manage the postmortem documents.
Sanitization: There are no instructions provided for sanitizing or escaping the content retrieved from external logs or chats before it is included in markdown files.
[COMMAND_EXECUTION]: The skill uses the exec tool to perform local file system operations and searches within the project directory.
Evidence: The skill provides examples for using standard utilities like ls, cat, and grep to interact with files inside the postmortems/ folder.

write-a-postmortem