The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of git and the GitHub CLI (gh) to manage repository status, create labels, and orchestrate issue creation. These operations are scoped to the current repository context and are necessary for the skill's primary function.
[EXTERNAL_DOWNLOADS]: The skill requires the external issue-creator skill to be installed. The instructions point to the author's official GitHub repository (github.com/luongnv89/idd). As this is a vendor-owned resource by the skill author, it is considered a legitimate dependency for the ecosystem.
[DYNAMIC_EXECUTION]: A bundled Python script, scripts/render_dashboard.py, is executed locally to render the epic dashboard. The script is deterministic, uses standard libraries only, and includes logic to escape untrusted content, mitigating risks associated with malformed plan data.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (plan files and user conversations). It mitigates potential indirect prompt injection by:
Utilizing explicit boundary markers () for dashboard regions.
Implementing a 'Prompt Injection Boundary' policy that treats all plan data as content, never as commands.
Using jq to read values from JSON worklists into shell variables at runtime, which prevents the shell from parsing untrusted text as syntax.
[SAFE]: The skill demonstrates a high level of security awareness. It includes a mandatory preflight phase to verify tool permissions and API budgets, and it employs 'verify-by-re-read' patterns to ensure all tracker mutations match intended states without relying solely on command exit codes.

plan-to-issues