The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies and mitigates risks associated with processing untrusted data from product briefs, specs, and tracker issues.
Ingestion points: The {source} file and tracker data (issue bodies and comments) loaded during the workflow.
Boundary markers: Explicit instructions in references/agentic-setup.md command the agent to treat repo and tracker content as data only, ignoring and reporting any embedded instructions like "ignore previous instructions."
Capability inventory: Tracker modification via update-issue, comment-issue, and the om-prepare-issue tool.
Sanitization: Mandatory regex validation and quoting for all externally-sourced values used in shell or path interpolation.
Human Review: A required confirmation step (HARD-GATE) ensures no issues are filed without user approval of the exact content.
[COMMAND_EXECUTION]: The skill interacts with the local environment using predefined companion tools.
Safety: Execution is limited to tools in the locally installed collection (e.g., om-prepare-issue). Input validation rules are provided to prevent command injection by sanitizing IDs and paths from the tracker or source files.

om-backlog