The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill explicitly instructs the agent to perform dynamic code generation and execution in references/validation.md. Specifically, it directs the agent to "Run a 10 to 30 line script and capture the output" to validate the behavior of code, libraries, or APIs during the planning phase.
[COMMAND_EXECUTION]: The workflow requires the execution of local shell commands and scripts. SKILL.md (Step 9) and references/plan.md (Definition of Done) mandate running a script located at <plugin root>/scripts/prose-lint.py. Furthermore, references/cross-validation.md (Step 7) involves running git log and grep to sweep the codebase for conflicts.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a wide variety of untrusted data from the local codebase, including product specifications, architecture documentation, and source code. This data is then used to influence the agent's planning and validation steps. Because the agent also has script execution capabilities, malicious instructions hidden in the codebase (e.g., in markdown comments or documentation) could theoretically be executed or lead to exfiltration, although no explicit exfiltration patterns were detected.

blueprint