The Agent Skills Directory

[COMMAND_EXECUTION]: The skill interpolates user-provided arguments $0 (rule name) and $1 (glob pattern) directly into shell commands. For example, pnpm exec eslint --rule '@sentry/scraps/$0: error' "$1" is used to count and fix violations. This presents a risk of command injection if the input is not sanitized, as shell metacharacters (e.g., ;, &&, |, or backticks) could be used to execute arbitrary commands.
[DYNAMIC_EXECUTION]: For large-scale rollouts exceeding 500 violations, the skill suggests that the agent write and execute temporary codemods or scripts using jscodeshift or @typescript-eslint/typescript-estree. This involves the runtime generation and execution of code logic, which can be an attack vector if the script generation is influenced by malicious input.
[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves reading and modifying source code files. This creates an ingestion surface where a malicious file in the codebase could contain instructions designed to influence the agent's behavior during the linting and fixing process.
Ingestion points: Source files matching the glob pattern provided in the $1 argument are read by the agent (referenced in SKILL.md).
Boundary markers: No explicit boundary markers or "ignore instructions" warnings are used when processing the content of these files.
Capability inventory: The skill possesses capabilities to execute shell commands (eslint, prek) and perform filesystem writes (applying lint fixes) across the codebase.
Sanitization: The skill instructions do not specify any sanitization or validation of the file content before the agent attempts to process and fix violations.

lint-fix