The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is instructed to read configuration data from ~/.comet/config.yaml. Accessing files in the user's home directory is a sensitive operation that can lead to the exposure of tool configurations or potentially sensitive tokens if they are stored within the same file.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from multiple external and project-level sources to drive its logic, which could be exploited to influence the agent's behavior if those files contain malicious instructions.
Ingestion points: The skill reads text returned by the comet task CLI command, as well as the contents of proposal.md, design.md, and tasks.md from the project directory.
Boundary markers: There are no instructions to treat the ingested text as untrusted data or to use delimiters to prevent the agent from following instructions embedded within these files.
Capability inventory: The skill possesses the ability to execute shell commands via the Comet and OpenSpec CLIs, run project build tools (mvn, npm), perform file system modifications (git commits), and load additional skills (comet-verify, comet-archive).
Sanitization: No sanitization, validation, or escaping logic is defined for the data ingested from tasks or project documentation.

comet-hotfix