Skip to content

workflow-recipes.md

Latest commit

 

History

History
98 lines (74 loc) · 2.43 KB

workflow-recipes.md

File metadata and controls

98 lines (74 loc) · 2.43 KB

Workflow Recipes

These are opinionated runbooks for common jobs-to-be-done.

Recipe 1: "I need a fast confidence check"

make check
./scripts/dp review --json

Success criteria:

  • make check passes
  • review returns "ok": true

Use when:

  • You changed code and want local confidence before deeper validation.

Recipe 2: "I need release-grade confidence"

make check
./scripts/dp verify --json
./scripts/dp enforce pre-commit --policy dp-policy.json --json
./scripts/dp enforce pre-push --policy dp-policy.json --json

Success criteria:

  • All commands return exit 0
  • JSON outputs include "ok": true

Use when:

  • You are about to merge or push important changes.

Recipe 3: "I need to formalize NL requirements"

  1. Capture concise requirements language.
  2. Run spec-agent workflow in your integration surface.
  3. Confirm generated artifacts include required level:
  • Types
  • Invariants
  • Contracts
  • Full proofs (when needed)
  1. Run verification gates.

Read more:

  • docs/spec/SPEC-20-typed-signatures.md
  • docs/spec/SPEC-22-proof-protocol.md
  • docs/spec/SPEC-27-fallback-extraction.md

Recipe 4: "A gate failed and I need triage"

  1. Re-run the exact failing command.
  2. Capture full output (no screenshots of terminal corners).
  3. Classify failure:
  • Build/toolchain
  • Test regression
  • Governance policy failure
  • Environment issue
  1. Apply targeted fix.
  2. Re-run complete gate sequence.

Then use:

Recipe 5: "I need to hand off cleanly"

Before handoff:

  1. git status
  2. make check
  3. ./scripts/dp verify --json
  4. Commit with clear scope.
  5. Push.
  6. Ensure branch is up to date with remote.

Because nothing says "teamwork" like not leaving broken local-only miracles.

Recipe 6: "I need confidence in Claude adapter behavior"

make claude-adapter-gate

Success criteria:

  • VG-CLAUDE-ADAPTER-E2E-001 passes
  • At least two scenario tests execute (guardrail against zero-test filter drift)

Use when:

  • You changed adapters/claude_code/*
  • You changed complexity signals, context externalization, or adapter mode logic
  • You want evidence for observe/orient/decide/act behavior, not just unit plumbing

Read:

  • /Users/rand/src/loop/docs/user-guide/claude-code-adapter.md

Command lookup companion:

  • /Users/rand/src/loop/docs/reference/command-reference.md