The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is primarily designed to analyze untrusted data provided by users, such as console error logs, stack traces, and symptom descriptions. Because the agent is equipped with powerful capabilities like Bash, Edit, and Write, a maliciously crafted log could attempt to influence the agent's behavior.
Ingestion points: SKILL.md (Workflow Step 1) and references/agent-guidance.md (Phase 1) explicitly direct the agent to gather verbatim error text and reproduction steps from the user.
Boundary markers: The instructions do not provide explicit delimiters or "ignore instructions within data" warnings to the agent when processing these logs.
Capability inventory: The skill allows the agent to use Bash, Write, Edit, and WebFetch tools, providing a surface for command execution or file modification if an injection is successful.
Sanitization: No sanitization or validation logic is defined for the ingested error content.
[COMMAND_EXECUTION]: The skill requires the execution of shell commands for reproduction and testing purposes.
Evidence: The references/agent-guidance.md file instructs the agent to run a headless playtest script located in a sibling directory (node "${CLAUDE_PLUGIN_ROOT}/skills/phaser-playtest/scripts/playtest.mjs") and to utilize git bisect run npm run build for automated regression testing, which executes build scripts defined in the project's configuration.

phaser-debugger