The Agent Skills Directory

COMMAND_EXECUTION (LOW): The skill utilizes bash to execute commands like jq and cat for reading configuration and memory files within the .specweave/ directory.
Evidence: jq -r '.planning.deepInterview.enabled // false' .specweave/config.json and cat .specweave/skill-memories/architect.md.
PROMPT_INJECTION (LOW): The skill is susceptible to indirect prompt injection (Category 8) because it ingests data from local configuration and memory files that might be modified by an attacker.
Ingestion points: .specweave/config.json, .specweave/skill-memories/architect.md.
Boundary markers: Absent. The skill reads raw output from jq and cat directly into its context.
Capability inventory: The skill has Read, Write, and Edit permissions, and the ability to chain to other domain-specific skills (e.g., sw-backend, sw-infra).
Sanitization: Absent. The content of the memory/config files is not validated or escaped before being processed by the LLM.

architect