The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches service status information and health data from official Granola domains (status.granola.ai and api.granola.ai). These operations are consistent with the skill's purpose as an incident response tool for that service.
[COMMAND_EXECUTION]: The skill utilizes pgrep and pkill to monitor and restart the Granola application process. It also reads local application logs using the tail command. These actions are appropriate for the skill's stated role in troubleshooting and remediation.
[REMOTE_CODE_EXECUTION]: Automated scans flagged a command piping curl output to python3. However, analysis confirms this is a safe pattern: the command uses the -c flag to execute a hardcoded local script string to parse the JSON data retrieved by curl. No code is being downloaded or executed from the remote source.
[INDIRECT_PROMPT_INJECTION]: The skill has the capability to ingest untrusted data from external status APIs and local log files.
Ingestion points: External status JSON (via curl) and local application logs (via tail).
Boundary markers: No explicit delimiters or instructions are used to separate untrusted content from the agent's instructions.
Capability inventory: The skill has access to shell execution via Bash (pkill, pgrep, curl, tail), and file modification via Write and Edit tools.
Sanitization: While the external JSON is parsed using a structured Python script, the text content within the JSON (such as status descriptions) is not sanitized before being presented to the agent.

granola-incident-runbook