The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Node.js and Bash scripts to automate Linear interactions. It executes system binaries such as curl, jq, and sqlite3. The implementation follows security best practices by using argument arrays and passing sensitive configuration data via stdin to prevent exposure in the process table or log files.
[DATA_EXFILTRATION]: The skill performs network operations via curl to communicate with the vendor's API at staging.catalystcloud.dev. This communication is the primary intended function of the tool, used to proxy writes to Linear as the application actor, and targets an established vendor domain using secure per-host authentication.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The triage and reporting scripts (ask-triage.sh, human-blocked.sh) read issue titles, status fields, and human comments from a local SQLite replica of Linear data to present a ranked list of tasks to the agent.
Boundary markers: The information retrieved from tickets is provided to the agent as formatted reports or JSON without specific delimiters or instructions to ignore potentially adversarial content embedded in the ticket descriptions or comments.
Capability inventory: The agent is empowered to create new tickets, post threaded comments, and advance ticket states (e.g., to 'Done') based on the data it processes.
Sanitization: While the tool includes logic to prevent local file paths from being incorrectly posted as comment bodies, it does not sanitize the text content of the tickets themselves for instructions that might attempt to influence the agent's behavior.

ask