The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs several local command-line operations, including verifying forge authentication via gh auth status and executing a project-specific merge hook at .agentic-workflow/hooks/fullauto-merge.sh when the --fullauto mode is enabled.
[EXTERNAL_DOWNLOADS]: The skill facilitates the expansion of agent capabilities by suggesting the installation of additional tools from the author's ecosystem using the command npx skills add gtrabanco/agentic-workflow.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data sources like the ROADMAP.md file and git forge issues.
Ingestion points: Reads project documentation and issue descriptions from external providers.
Boundary markers: Explicitly mandates that once a run is founded, all decisions must be resolved from the committed SHIP_DECISIONS.md file, creating a trust boundary against instructions in implementation docs or issues.
Capability inventory: Includes subagent orchestration, shell hook execution, and package management via npx.
Sanitization: Specifically instructs the agent to use --body-file for forge interactions to avoid potential shell injection attacks via hand-escaped markdown content.
[DYNAMIC_EXECUTION]: Implements a tiered architecture where the conductor skill spawns subagents for implementation tasks using specific model overrides (e.g., Sonnet), ensuring that execution logic is separated from high-level orchestration.

ship-roadmap