The Agent Skills Directory

[COMMAND_EXECUTION]: The skill triggers the execution of various local utility and testing scripts, such as python3 fdk/tools/medic.py, python3 fdk/tools/ci-local.py, and bash harness/scripts/fresh-install-smoke.sh. These are used to validate code health and installation readiness before deployment.
[EXTERNAL_DOWNLOADS]: During the User Acceptance Testing (UAT) phase, the skill fetches and executes code from the project's remote repository using curl. This process targets well-known services like GitHub and GitLab to verify that the remote installation process is functioning correctly.
[DYNAMIC_EXECUTION]: The skill dynamically assembles and executes shell commands for Git and repository management tools like gh (GitHub) and glab (GitLab). Safety is maintained through a mandatory 'STOP' and 'DỪNG' checkpoint, ensuring no side-effects occur without explicit user confirmation.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from commit history and local logs to generate release notes and Pull/Merge Request descriptions.
Ingestion points: Git commit logs (git log) and session scratch-logs are read in Step 6 of the main workflow.
Boundary markers: Absent; the skill relies on instructional context to format the output.
Capability inventory: The agent has capabilities to execute shell commands, perform network operations (git push/PR creation), and write files.
Sanitization: No explicit sanitization or filtering is performed on the log data before it is processed by the model to generate descriptive text.

ship