The Agent Skills Directory

[DYNAMIC_EXECUTION]: The orchestrator generates and executes Node.js scripts via the node -e command. These scripts use string interpolation to inject data parsed from external product plans (e.g., spec-tech*.md), which could lead to code injection if the files are maliciously crafted. This pattern is used for tracking state and seeding tasks in stelow.json.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from spec-tech*.md files. This content defines the scope objective, acceptance criteria, and verify commands used for subagent delegation. The absence of strict boundary markers or sanitization for these interpolated fields creates a vulnerability surface.
[COMMAND_EXECUTION]: Employs shell commands such as git rev-parse, git diff, and grep to interact with the repository, monitor changes, and determine the file footprint of executed tasks.
[DATA_EXPOSURE]: Accesses and modifies project configuration in stelow.json and documentation files within the .stelow/ and docs/ directories. While these are project-specific files, the automated modification of state files based on external input requires caution.
[REMOTE_CODE_EXECUTION]: Dynamically resolves and loads the @calionauta/stelow package components using require() with computed paths. As these are vendor-provided resources from the skill's author, this represents standard functionality but involves dynamic loading techniques.

stelow-product-scope-executor