The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow relies on executing various shell commands via Bash for reproduction and verification tasks. This includes using curl to probe local API endpoints, docker compose exec for database inspection, redis-cli for queue status, and running project-specific build and test scripts such as npm run dev, tsc, lint, and test.
[DYNAMIC_CONTEXT_INJECTION]: The SKILL.md file contains !command`` placeholders in the <live_context> section. These execute git log and git status commands at skill load time to provide the agent with current repository information. These commands are benign and standard for development tools.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, potentially attacker-controlled sources, which presents a surface for indirect instructions.
Ingestion points: User-supplied arguments, Sentry issue descriptions and stack traces, PostHog event data, and browser console/network logs fetched during the reproduction phase.
Boundary markers: The instructions do not define strict delimiters or include "ignore embedded instructions" directives for the data retrieved from external monitoring tools or browser sessions.
Capability inventory: The skill has broad system access, including shell execution, file system modification (to apply bug fixes), and the authority to dispatch and configure autonomous sub-agents.
Sanitization: There are no explicit instructions for sanitizing or validating content retrieved from external APIs or browser environments before it is processed by the agent's logic.

debug