The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/runtime/common.mjs contains a runChild function that wraps node:child_process.spawn. This utility allows for the execution of arbitrary system commands.
Evidence: spawn(program, args, { stdio: ["pipe", "pipe", "pipe"], windowsHide: true }); in scripts/runtime/common.mjs.
Context: While this function is a generic utility and is not called by the integrity check script (scripts/run.mjs), its presence provides the skill with subprocess execution capabilities.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted project data, including code, configurations, and documentation, to generate architectural artifacts.
Ingestion points: As described in SKILL.md, the skill reads .devflow/index.md, project code, configuration files, and existing architectural documentation.
Boundary markers: The skill relies on natural language instructions to distinguish evidence from analysis but lacks explicit technical delimiters or sanitization for all external data sources.
Capability inventory: The skill has access to the file system (via node:fs/promises) and command execution capabilities (via runChild in the utility library).
Sanitization: The scripts/runtime/common.mjs file includes a resolveProjectPath function designed to prevent directory traversal attacks by ensuring all file access remains within the designated project root.

architecture