feat(planning): route new work by diff size and fog: one-shot, /planning:interview, or /planning:wayfind #6729

Activity

1. github-actions

   added

   needs-triageNot yet classified. Floor until a type and one priority tier are set.Not yet classified. Floor until a type and one priority tier are set.

   on Oct 10, 2026
2. kyle-sexton commented on Oct 10, 2026

   

   kyle-sexton

   on Oct 10, 2026

   MemberAuthor

   More actions

   Sources for the routing

   Matt Pocock, 2026-10-08: https://x.com/mattpocockuk/status/2108216899439894574 (the flow diagram is attached to this post)

   1. If the diff is tiny, DON'T grill. /grill-with-docs saves you from situations where the agent built the wrong thing. But if the thing it's building is tiny, that cost drops to nothing.

   2. Never start with /wayfinder. A failure mode of /wayfinder is that you start grilling and the solution turns out to be way simpler than you thought ... you've created the map and tickets, and you realise you don't need them. So if the expected diff is medium-large, always start with a /grill-with-docs session. Then, if planning starts to get out of hand, just say "/wayfinder let's turn this into a map".

   Dex Horthy (HumanLayer), 2026-10-09: https://x.com/dexhorthy/status/2108588513436582357

   if you're using heavy planning to do small stuff you're gonna feel slower. I predict more of these tools/skills will start to emphasize escape hatches for human or model to decide "oh wait this is small, let's just go ship" ... the "smart zone" gets a little bigger (line is closer to 200-400k now I think). At humanlayer we've been prototyping research -> implement or research -> short design -> implement workflows

   Additional acceptance criteria

   • Never start with /planning:wayfind. When the expected diff is medium or large, entry is always /planning:interview. Wayfind is reached only by escalating from an interview, never first. If /planning:wayfind is invoked cold, it suggests starting with an interview instead.
   • Escape hatch inside the interview. At any round boundary, either the operator or the model can say "this is small, ship it". The interview then stops asking, writes the decisions so far, and hands off to implementation.
   • The smart-zone threshold for the "bigger than you thought?" check comes from context-guard's zone signal. Don't hard-code a token count in the skill. The zone thresholds belong to context-guard.

   🤖 Generated with Claude Code

   https://claude.ai/code/session_01LGDKQK7XJf6kz6EZDzbKwM
3. github-actions commented on Oct 10, 2026

   

   github-actionsBot

   on Oct 10, 2026 – with GitHub Actions

   Contributor

   More actions

   Thanks for the report. Automated triage handed this issue to a maintainer.

   _{Automated triage. A maintainer reviews it.}
4. github-actions

   added

   needs-humanHuman-in-the-loop required; autonomous sessions must not resolve items carrying this.Human-in-the-loop required; autonomous sessions must not resolve items carrying this.

   on Oct 10, 2026
5. kyle-sexton commented on Oct 11, 2026

   

   kyle-sexton

   on Oct 11, 2026 – with Claude

   MemberAuthor

   More actions

   Draft PR: #7102 (branch feat/6729-route-by-diff-size-and-fog).

   • /planning:interview gets a "Routing by size and fog" section: a bigger-than-you-thought check at every round boundary (context-guard zone worsened, or two rounds running opened more questions than they closed) that recommends /planning:wayfind chart with the ledger, constraints, intake disclosure and glossary carried over; a "this is small, ship it" exit (open rows deferred, hand off to /implementation:implement); and the reason the entry gate is diff size and retry cost. Wayfind's chart step also takes those carried items. The tiny-diff gate and wayfind's "never first" gate were already in the skills.
   • Verified locally: affected-tests.sh --base origin/main --run (4 suites pass, including interview-defenses), markdownlint, fragment check.
   • Left for a human: marking the PR ready; the "two rounds running" bar is my default. The needs-human label is left for Kyle to clear.

   ---

   Generated by Claude Code
6. kyle-sexton commented on Oct 11, 2026

   

   kyle-sexton

   on Oct 11, 2026 – with Claude

   MemberAuthor

   More actions

   Handoff note for whoever picks this up next. The overnight Claude run has stopped, and Kyle is moving issue work to Cursor.

   • Branch: feat/6729-route-by-diff-size-and-fog
   • PR: feat(planning): route by size and fog, with a mid-interview check and a ship-it exit #7102 (draft)
   • State: Verifier FIX-NEEDED (SKILL.md 510 lines > 500 cap; disclosure not persisted for wayfind; ship-it vs Step 3 confirm), fixed in c5c4bf3 (routing text moved to context/routing.md, SKILL.md 497 lines; Starting point section in ledger; ship-it keeps Step 3 confirm and skips the assumption sweep). Merged with main it is 499 lines, one under the cap. Kyle: Stance :111/:436 still treat one opening round as the wayfind signal (attended digest). "Two rounds running" fog check is a default for Kyle; section placed before does-NOT-do to avoid attended digest pins; evals not run
   • Next step: Re-verify the latest PR head locally (scripts/affected-tests.sh --base origin/main --run plus the checks listed in the PR body), finish anything listed as open, then review.

   Where a branch is listed it is pushed, so it can be taken over as is. A draft PR shows a red ci-status by design until it is marked ready.

   ---

   Generated by Claude Code
7. kyle-sexton

   removed

   needs-triageNot yet classified. Floor until a type and one priority tier are set.Not yet classified. Floor until a type and one priority tier are set.

   on Oct 12, 2026
8. kyle-sexton

   added the

   Feature

   issue type on Oct 12, 2026
9. kyle-sexton commented on Oct 13, 2026

   

   kyle-sexton

   on Oct 13, 2026

   MemberAuthor

   More actions

   Decision recorded. The owner approved applying independently verified recommendations on 2026-10-12.

   • How many consecutive rounds that open more questions than they close should trigger the wayfind recommendation? → One round (match the pinned Stance text, no digest change)
     Basis: judgment; PR feat(planning): route by size and fog, with a mid-interview check and a ship-it exit #7102 body ('two rounds running bar is my choice'; Stance 'Question budget' ~line 111 says a single round)
   • Run the interview/wayfind evals before marking feat(planning): route by size and fog, with a mid-interview check and a ship-it exit #7102 ready? → Yes
     Basis: PR feat(planning): route by size and fog, with a mid-interview check and a ship-it exit #7102 Verification ('Not run: interview/wayfind evals'); AGENTS.md Code Review Rules (skill criteria, eval evidence)

   Draft PR #7102 adds a mid-interview 'bigger than you thought?' check, a ship-it exit, and the wayfind handoff in a new context/routing.md. Left open: the PR's 'two rounds running' trigger conflicts with pinned Stance text that treats one round as the signal.

   Verification notes:

   • Minor: Q2's basis overstates the rule; AGENTS.md requires eval evidence for a cut to an existing skill, and feat(planning): route by size and fog, with a mid-interview check and a ship-it exit #7102 is mostly additive. Running the evals is still a reasonable pre-ready step.

   Verified by an independent fresh-context check (confirmed, risk low): gh issue view 6729 incl. owner source comment and 2026-10-11 handoff; PR #7102 draft, MERGEABLE, head c5c4bf3; body lines on 'two rounds running' default, Stance ~111, evals not run; origin/main SKILL.md:109,111 (single-round ballooning signal), :438; git merge-tree origin/main + #7102: clean; merged SKILL.md 499 lines (assumption true)

   Admission to an agent lane stays with the work-class rules and owner ratification.
10. kyle-sexton

    closed this as completedin #7102on Oct 13, 2026