The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted build, CI, and deployment logs, which constitutes an indirect prompt injection attack surface.
Ingestion points: Untrusted data is ingested when the agent reads build logs, CI step outputs, and deployment logs as described in SKILL.md and rules/diagnosis-resume-and-safety.md.
Boundary markers: The skill includes strong instructions in SKILL.md and rules/diagnosis-resume-and-safety.md to treat external data as evidence only and to ignore instructions or commands embedded in that data.
Capability inventory: The skill utilizes Bash for local command execution and file editing tools (Read, Write, and Edit) as defined in SKILL.md.
Sanitization: The skill relies on instructional constraints and a phased workflow requiring user confirmation for fix options, rather than programmatic sanitization of log content.

deploy-fix