The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external Sentry issue reports, such as stacktraces, titles, and descriptions, which could contain malicious instructions designed to influence the agent's behavior during code modification or command execution.
Ingestion points: Data enters the agent's context through the mcp__sentry__get_sentry_resource tool as described in section 4a of SKILL.md.
Boundary markers: The instructions lack explicit delimiters or specific directives for the agent to ignore instructions embedded within the Sentry issue data.
Capability inventory: The skill has the capability to modify the local codebase, execute shell commands for quality gates (e.g., PHPStan, Pint), and perform git commits as seen in sections 4e and 4f of SKILL.md.
Sanitization: There is no mention of sanitizing or validating the Sentry-provided content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill performs command execution to run project-specific quality gates, including static analysis tools like PHPStan, code style fixers like Pint, and general test suites. These operations are intended for local validation and require explicit user approval before execution.

sentry-triage