The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands to verify the environment and determine the project root. Evidence: pwd, git rev-parse --show-toplevel.- [CREDENTIALS_UNSAFE]: The skill instructs the agent to open and read sensitive environment configuration files to extract values for planning. Evidence: Instructions to access .env and other configuration files like pake.json or tauri.conf.json to 'lift the live value'.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by design, as it incorporates and prioritizes instructions found in unverified local files.
Ingestion points: Reads instructions from AGENTS.md, CLAUDE.md, .claude/rules/*.md, and user-provided memory paths.
Boundary markers: Absent; the skill is instructed to follow rules found in these files and surface contradictions rather than ignoring them.
Capability inventory: Shell command execution (pwd, git) and broad filesystem read access across the project directory.
Sanitization: No sanitization or validation is performed on the content ingested from the local instruction files.

think