Skip to content

[codex] Recalibrate profile recipe around three-loop use#75

Closed
nahiyankhan wants to merge 2 commits into
mainfrom
recalibrate-profile-recipe
Closed

[codex] Recalibrate profile recipe around three-loop use#75
nahiyankhan wants to merge 2 commits into
mainfrom
recalibrate-profile-recipe

Conversation

@nahiyankhan
Copy link
Copy Markdown
Collaborator

🤖 Raised by Codex on behalf of Nahiyan.

Summary

  • Recalibrates the profile recipe around rendered-product grounding, token extraction, and downstream generator/reviewer/comparator loops.
  • Updates the template and schema guidance so profile output is more portable and less biased by placeholder defaults.
  • Adds a changeset for the published skill-bundle-facing behavior.

Notes

There was a prior closed PR for this branch; this opens a fresh draft because the local branch still carries a unique unapplied patch.

Validation

  • Not run in this pass; this PR packages existing local branch work.

nahiyankhan and others added 2 commits April 26, 2026 16:27
…table use

Rewrite the profile recipe so the host agent grounds in the rendered
product before tokens, frames the expression for portability across
web/iOS/Android, and writes decisions in operationalizable form
(pattern + how-to-spot-a-violation). Promote composition as a
first-class kind of decision alongside visual phenomenon. Add a
coverage-audit practice for roles[] so absences are stated as
decisions instead of left silent for downstream generators to invent
around. Strip biasing defaults from expression.template.md and shift
role naming guidance toward cross-platform archetypes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* main: (50 commits)
  Refine Ghost docs IA and copy
  Rename expression to fingerprint
  Clarify agent-facing Ghost guidance
  Make expression.md an authored-only contract; add verify-profile gate
  dogfood(ghost-ui): attempt-7 — profile via bounded survey summary
  Add ghost-expression survey summarize for bounded profile context
  dogfood(cash-ios-moneybot): three-stage scan
  Upgrade scans to map/v2 + survey/v2 with implemented-surface evidence
  test: update writer test for # Local References rename
  dogfood(dsgn-playground): three-stage scan
  Make expression.md portable across projects
  Rename bucket evidence to survey
  Improve expression review and generation steering
  Tighten expression scans and terminal context
  recipe(survey): require Tailwind class-atom pass
  dogfood(ghost-ui): attempt-3 — three-stage scan post schema-trim
  dogfood(managerbot): three-stage scan of square-web/apps/managerbot
  schema(expression): drop signature section and distinctiveTraits field
  invariant: drop signature from the body partition
  dogfood(ghost-ui): canonical stage-2 bucket at package root
  ...

# Conflicts:
#	packages/ghost-drift/src/skill-bundle/references/profile.md
#	packages/ghost-drift/src/skill-bundle/references/schema.md
#	packages/ghost-fingerprint/src/skill-bundle/assets/fingerprint.template.md
Copy link
Copy Markdown
Collaborator Author

🤖 Follow-up: I merged current main into this branch to resolve the stale-branch conflicts. The PR is now mergeable. Validation from the hooks on the merge/push:

  • pnpm check
  • pnpm build
  • pnpm test (405 tests passed)

@nahiyankhan
Copy link
Copy Markdown
Collaborator Author

🤖 Closing this draft PR because it came from an old/stale branch and should not stay in the active queue.

@nahiyankhan nahiyankhan closed this May 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant