The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local CLI commands and scripts to manage tasks and verify system state.
Evidence: SKILL.md requires trellis on the system path and executes trellis -v for version detection.
Evidence: SKILL.md invokes task.py create and task.py start to initialize dedicated repair tasks.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its core logic is driven by the contents of an external report.
Ingestion points: SKILL.md reads and parses findings from WORKFLOW_QUESTIONS.md in Step 1 and Step 2.
Boundary markers: The skill validates the report protocol (workflow-scan-repair-v2) and frontmatter but lacks delimiters to prevent the agent from following instructions potentially embedded within the findings.
Capability inventory: The skill has write access to docs/workflows/新项目开发工作流/ and the ability to execute task management scripts.
Sanitization: There is no explicit sanitization of finding descriptions or evidence parsed from the report before they are used to generate a correction plan.

workflow-repair