RIG-401 feat: implementation

[ArLeyar]

Summary

Pipeline engineer task 20260407-039.

Linear: RIG-401

[ArLeyar]

Code review

Found 1 issue:

1. Stages without explicit runtime default to claude-code, so honeyjourney.yaml's analyst/deployer stages will generate skill: "python" — but no /python skill exists in Claude Code. The skill lookup will fail or stall on every analyst/deployer task in the honeyjourney pipeline.

werma/engine/src/pipeline/prompt.rs

Lines 110 to 117 in eb438ba

	} else {
	"## FIRST: Invoke the Code Review skill\n\
	Before starting the review, invoke the `/code-review` skill using the Skill tool \
	(skill: \"code-review:code-review\"). This loads the full review checklist and \
	standards you MUST follow.\n\n\
	## Review Protocol\n\
	1. Invoke `/code-review` skill (Skill tool, skill: \"code-review:code-review\")"
	.to_string()

Root config that triggers it:

werma/engine/pipelines/honeyjourney.yaml

Lines 6 to 9 in eb438ba

	templates:
	verdict_instruction: "Output your verdict on the LAST line of your response"
	nit_threshold: "0"
	language: python

Fix: add a runtime guard in compute_derived_vars so skill invocation is only generated for languages with a known installed skill (e.g. ["rust"] allowlist), or set an explicit non-claude runtime on the honeyjourney analyst/deployer stages.

🤖 Generated with Claude Code

_{- If this code review was useful, please react with 👍. Otherwise, react with 👎.}