The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes untrusted codebase data, including source code, documentation, and configuration files. This creates a surface where malicious instructions embedded in the analyzed codebase (e.g., within comments or README files) could attempt to influence the agent's behavior during the analysis.
Ingestion points: Multiple steps in SKILL.md read from {codebase_path} using find, grep, and cat (Steps 1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13).
Boundary markers: The skill does not explicitly use delimiters or instructions to ignore embedded prompts in the analyzed content.
Capability inventory: The skill executes shell commands (find, grep, cat), invokes the TypeScript compiler (npx tsc), and runs local Python scripts.
Sanitization: No explicit sanitization or filtering of the ingested file content is mentioned before processing.
[COMMAND_EXECUTION]: The skill uses various shell commands to inspect the codebase structure and content.
Evidence: Steps 1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, and 13 in SKILL.md use commands like find, grep, cat, and wc to analyze the filesystem and file contents.
[DYNAMIC_EXECUTION]: The skill performs runtime analysis using compilers and helper scripts.
Evidence: Step 11 in SKILL.md executes npx tsc --noEmit to identify TypeScript errors. Steps 3 and 4 call local Python scripts (extract_tech_stack.py and validate_metrics.py) located in .claude/skills/bmad-commands/scripts/ to perform automated data extraction and metric validation.

analyze-architecture