Skip to content

feat(planning): chart a data-sharing item for a selected service that receives user data - #16551

Open
kyle-sexton wants to merge 2 commits into
mainfrom
feat/wayfind-data-sharing-item
Open

kyle-sexton wants to merge 2 commits into
mainfrom
feat/wayfind-data-sharing-item

Conversation

@kyle-sexton

@kyle-sexton kyle-sexton commented Nov 16, 2026 •

Copy link
Copy Markdown
Member

No related issue: owner decision this session; no tracker item covers it.

Summary

/planning:wayfind gains one rule: when a decision selects an outside service that will receive
user data, the map charts one interview item for that service covering which fields cross, under
what agreement, and, in a regulated repository, what must be in place before build (for example
whether a business associate agreement is needed). Its resolution feeds the PRD. A service that
receives no user data gets no item. Before this change nothing in the skill raised that question
(grep for user data, business associate and regulated across plugins/planning found no
such rule).

A wayfind: prototype item that proves an outside vendor does something now routes to
/prototype:pressure-test in its vendor-fit mode.

Fix

  • plugins/planning/skills/wayfind/SKILL.md: work step 6 holds the data-sharing item rule; chart
    step 4 applies it to a settled answer that already selects such a service; the work-step-5
    routing table names the vendor-fit mode of pressure-test.
  • plugins/planning/skills/wayfind/evals/evals.json: case 28 adds an expectation that the vendor
    check runs through /prototype:pressure-test in a vendor-fit mode, asserted on the skill and the
    vendor-fit intent, so a slightly different final mode name still passes. New case 33 pairs a
    service that receives patient data (one data-sharing item expected) with one that receives none
    (no item expected).
  • plugins/planning/evals/wayfind-prototype-files-on-dossier-folder-and-safe-push/: the
    plugin-eval mirror of case 28 gets the same expected outcome plus a vendor-fit-route grader and
    its pass/fail samples.
  • .changes/planning/: a minor fragment.

session: AI Hero

Verification

  • scripts/check-changed-skills.sh origin/main: CHECK-SKILL wayfind: PASS — 0 errors, 6 warning(s) (five pre-existing; one notes the body grew by 8 lines).
  • check-jsonschema --schemafile plugins/skill-quality/reference/evals.schema.json plugins/planning/skills/wayfind/evals/evals.json: ok -- validation done.
  • validate-cases.py plugins/planning/evals: no FAIL; the changed case shows only the
    judge-calibration warnings every case in the suite carries.
  • scripts/check-changelog-fragments.sh --check and --check-required origin/main: pass.
  • prettier and markdownlint-cli2 on the changed files: clean. Pre-commit and pre-push hooks passed.
  • claude plugin eval, run locally on head 0542551f after the branch took main:
    • Case 33 (data-sharing item), converted from evals.json in a scratch copy, 3 runs per arm:
      VALID. With the plugin 15 of 15 grader checks passed; without it 14 of 15. One case is too few
      for the noise report to call a gain, the baseline is near the ceiling, and these graders are
      uncalibrated, so no gain is claimed.
    • Case 28 (wayfind-prototype-files-on-dossier-folder-and-safe-push): the new
      vendor-fit-route judge grader calibrated at 12 of 12 against its samples. The case itself
      returned INVALID on two runs (2 permission denials in the traces): the without-plugin arm
      tries to Read the plugin's SKILL.md and is denied. The case's prompt on main already asks
      the agent to read skill files at their exact path, so this is not caused by this change. No
      grader counts are quoted for an invalid run; the case has no valid result yet.

Related

https://claude.ai/code/session_01MWmASNT5Knyq99gphnV6B2

🤖 Generated with Claude Code

… receives user data

/planning:wayfind work step 6 now charts one interview item when a
resolution selects an outside service that will receive user data:
which fields cross, under what agreement, and, in a regulated
repository, what must be in place before build (for example a business
associate agreement). One item per service; none when no user data
crosses. Its resolution feeds the PRD. Chart step 4 applies the same
rule to a settled answer that already selects such a service.

A prototype item that proves an outside vendor does something routes
to /prototype:pressure-test in its vendor-fit mode. Eval case 28 and
its plugin-eval mirror assert that route by skill and intent, not an
exact flag; new case 33 pairs a service that receives user data with
one that receives none.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MWmASNT5Knyq99gphnV6B2
@kyle-sexton
kyle-sexton marked this pull request as ready for review November 16, 2026 15:42
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 16, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-16T15:44:27.831821Z a6b7137 Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a6b7137782

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

@github-actions

Copy link
Copy Markdown
Contributor

Lane run report

Field Value
Activity pr-refine/fix-docs
Model rung 0
Leg handoff
Run job success
Activity step success
Commits pushed none (read lane)
Run 43899400691

Failed GitHub API calls

Method Path Status
GET /repos/{repo}/git/ref/tags/* 404

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

claude-review: 1 inline finding(s) on a6b7137782d8d96cdbaaea6bad4da224a53034af; the summary is in this lane's pull request comment.

claude-lane: claude-review

@github-actions

github-actions Bot commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

claude-review needs no new review of 0542551f9407cc9eeced7fa8c3fd9f35b7b42e5d: no file of this pull request changed since the review of a6b7137.

Code review: PR #16551

I read all 6 changed files in .claude-lane/pr.diff. No issue closes this PR (the owner decided this in session), so there are no acceptance criteria to check against.

What changes

  • /planning:wayfind gains a data-sharing interview item. When a decision picks an outside service that will receive user data, the map charts one item for that service.
    • The rule sits in work step 6, and chart step 4 points to it.
  • The prototype routing row now names a vendor-fit mode of /prototype:pressure-test.
  • Eval case 28 and its plugin-eval mirror gain a vendor-fit expectation, with a new grader and its samples.
  • New eval case 33 covers a service that receives patient data and one that receives none.
  • A minor changelog fragment is added. That level fits an added capability.

Consistency checks

  • Case id 33 follows id 32.
  • Both cross-references in the new text exist in SKILL.md:
    • "Readiness" (L109)
    • step 8 closure (L335)
  • The new grader's pass and fail samples match its PASS and FAIL wording.

Finding (inline): the new routing names a pressure-test vendor-fit mode that doesn't exist on this base. The pressure-test skill offers only [observe] [--medium <value>] [scope]. The PR says the mode lands in a separate PR, but nothing here enforces that it merges first. Details are in the inline comment.

I ran none of the PR's listed verification commands myself. The skill-check, schema validation and changelog checks are the author's claims and are unverified.

Verifier dropped 0 of 1 candidates.

@github-actions

github-actions Bot commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

claude-security-review needs no new review of 0542551f9407cc9eeced7fa8c3fd9f35b7b42e5d: no file of this pull request changed since the review of a6b7137.

No security issues found. I read every hunk of all 6 changed files.

The change only adds instructions to /planning:wayfind:

  • Data-sharing interview item (new rule). When a decision selects an outside service that receives user data, the skill charts one interview item for it, covering which fields cross, under what agreement, and, in a regulated repository, whether a business associate agreement is needed.
  • Routing hint. A prototype item that proves an outside vendor does something routes to /prototype:pressure-test in its vendor-fit mode.
  • Evals and changelog. The rest is eval cases, an LLM grader, its sample answers, and a changelog fragment.

Instruction-surface lens. The diff deletes, narrows or softens no standing instruction:

  • The SKILL.md hunks only add text.
  • The routing-table row keeps "Its files go in the item's folder on the dossier branch" and the human-gated mode.
  • The task row's rule that a resolution records "names and locations of what was set up, never the values" is unchanged.
  • The new rule adds a privacy check. It removes none.

Nothing else here is security-relevant:

  • No plugin manifest, workflow, hook, script, trust control, token request, kill switch or merge hold changed.
  • None of the changed files matches a TRUSTED PATHS glob.
  • The PR context file has no embedded directives aimed at this lane.

Spend: the new eval case (id 33) and the new grader slightly increase prepaid eval-run usage when the suite runs.

critical-findings: 0

@kyle-sexton kyle-sexton added the do-not-merge Hard merge gate: do not merge while applied. label Nov 16, 2026
@kyle-sexton

Copy link
Copy Markdown
Member Author

Held: this change routes vendor-fit questions to a vendor-fit mode of /prototype:pressure-test that #16554 adds. Released when #16554 is merged and this branch is updated from main, so the route points at a mode that exists.

This branch has not been deployed

No deployments
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