The Agent Skills Directory

[SAFE]: The skill does not contain any malicious obfuscation, persistence mechanisms, or unauthorized privilege escalation attempts. The instructions focus on standard development workflows and follow-up procedures.
[COMMAND_EXECUTION]: The instructions permit the agent to execute shell commands for linting, testing, and verification during the implementation process.
Evidence: Step 2.4 and Step 3.1 explicitly mention running commands via the shell to verify code changes.
Safety Controls: The skill includes a specific "Nguyên tắc" (Principles) section that forbids dangerous operations such as rm -rf, DROP TABLE, git push --force, and recursive deletions.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it is designed to ingest and execute instructions from a previously generated implementation plan.
Ingestion points: The skill reads the "PLAN document" or summaries provided as output from the skill-plan process (SKILL.md).
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined to wrap the ingested plan content.
Capability inventory: The skill allows file system modifications and shell command execution.
Sanitization: No automatic sanitization of the plan is performed; however, the agent is instructed to use the AskQuestion tool and stop execution if the plan contains errors or unexpected instructions.

skill-implement