The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various CLI tools including openspec config list to verify workspace configuration and which sem followed by sem impact <token> for dependency analysis. It also uses python3 to generate ISO 8601 timestamps for artifact metadata. These commands use static paths and well-defined arguments relevant to the skill's purpose.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted ticket or feature request data which is then passed into prompt templates for sub-agents to generate research questions, design documents, and task plans. While the skill explicitly manages context isolation between phases, it lack explicit delimiters or sanitization for the untrusted input within sub-agent prompts. 1. Ingestion points: The initial ticket description provided by the user in SKILL.md Step 1. 2. Boundary markers: Absent; ticket content is interpolated into sub-agent prompts without unique delimiters or 'ignore' instructions. 3. Capability inventory: The skill has access to subprocess execution (openspec, sem, python3), file-write and edit capabilities (Write and Edit tools), and sub-agent spawning (Agent tool) as seen in SKILL.md Steps 10, 14, 25, and 37. 4. Sanitization: Absent; the raw ticket text is used to drive the workflow.
[DYNAMIC_EXECUTION]: The skill generates and executes short Python scripts via python3 -c to generate timestamps for the trace.md and design.md files in SKILL.md Steps 25 and 31. These are based on fixed templates and do not incorporate untrusted input into the command string.

accelint-qrspi-propose