The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file contains a 'Memory Protocol' section that explicitly directs the agent to execute the shell command cat .claude/context/memory/learnings.md. This instructs the agent to use its command execution capabilities to read specific local files as part of its operational flow.
[PROMPT_INJECTION]: The file commands/svelte-expert.md contains strong directives ('follow it exactly as presented to you') intended to enforce the skill's instructions. Additionally, the skill is susceptible to indirect prompt injection because it analyzes untrusted user-provided code while having access to powerful tools like Bash and Write without providing security boundaries for that code.
Ingestion points: User-provided source code files are analyzed for best practice compliance as defined in SKILL.md.
Boundary markers: The instructions do not define any delimiters or ignore-instructions for the code being reviewed.
Capability inventory: The skill is configured with the Read, Write, Edit, Bash, Grep, and Glob tools.
Sanitization: There is no evidence of input validation, escaping, or filtering of the content within the reviewed code files.

svelte-expert