The Agent Skills Directory

[SAFE]: The skill provides a legitimate project management workflow for agents. All instructions and scripts are consistent with the stated purpose of tracking session goals.
[COMMAND_EXECUTION]: The skill uses a local script (./scripts/goal.sh) to perform file operations and validation. The script is written defensively, using printf placeholders to prevent shell command injection from arbitrary user-provided goal titles and reason strings. It does not require or request elevated privileges.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-defined content, creating a potential surface for indirect prompt injection.
Ingestion points: User-supplied goal titles and item descriptions are stored in markdown files under .context/goals/ and processed by the tracking scripts.
Boundary markers: The skill uses standardized markdown headers and HTML comment blocks in its goal template to structure data, though it does not explicitly instruct the agent to ignore instructions embedded within the user data.
Capability inventory: The skill's capabilities are restricted to writing tracking files and executing local project-specific scripts. No network access or sensitive file access is permitted.
Sanitization: Titles are sanitized for use in filenames via a slugify function, and file content is written using printf %s to avoid unintended shell expansion or manipulation. Structural validation is enforced via the check subcommand.
[DATA_EXPOSURE]: All data handled by the skill remains local to the project repository's .context directory. No credentials, environment variables, or other sensitive system data are accessed or exposed.

goal-tracker