The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external codebase files, which constitutes a surface for indirect prompt injection. Maliciously crafted code or comments within the analyzed project could attempt to manipulate the architectural findings or the resulting implementation plans.
Ingestion points: The skill reads directory structures and file contents during the 'Sweep' and 'Deep Dive' analysis phases as defined in SKILL.md.
Boundary markers: The instructions for the subagents provide specific architectural lenses but do not include explicit delimiters or warnings to ignore instructions embedded within the codebase being analyzed.
Capability inventory: The skill has the ability to write files to the project repository, execute git commands (add, commit, push), and invoke the superpowers:writing-plans skill for code restructuring.
Sanitization: Content from the codebase is passed to analysis agents without specific filtering or sanitization of potential natural language instructions.
[COMMAND_EXECUTION]: The skill relies on a local helper script architect.sh to manage project configuration and state. This script executes shell commands including git for project detection and jq for JSON data processing. Analysis of the script shows that it uses these tools according to best practices, such as using jq arguments to handle variables safely.

architect