feat: planning modes + the Trace model (Tasks replaced by Traces) - #2
Open
xiduzo wants to merge 11 commits into
Open
feat: planning modes + the Trace model (Tasks replaced by Traces)#2xiduzo wants to merge 11 commits into
xiduzo wants to merge 11 commits into
Conversation
Add skills/references/planning-mode.md (resolve block, flow deltas, escalation rules). wtf.write-feature and wtf.epic-to-features resolve the mode from an argument or .wtf/config.json; flow mode batches questions, fans drafting out to subagents, and reviews once. wtf.setup asks the mode and records the planning config key. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Design doc for the tracer-bullet work model: Epic (horizon) -> Feature (step, 1..n co-related stories) -> Trace (one story, end-to-end, spine-first sequencing). Records the four decisions, skill impact, phased rollout, and open questions. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A grilling session on 2026-08-11 resolved all four open questions. A Trace now claims one story plus a Scenario Claim that partitions the story's scenarios. Re-aim runs through wtf.refine (interactive or headless, grow-only). Scenarios stay canonical in the Feature body with ephemeral .feature projection at verify time. Delivery is staged or trunk via .wtf/config.json. CONTEXT.md adds the glossary. ADR 0001 records the replacement decision. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Document guided/flow planning modes (config key, per-invocation override, flow behavior of epic-to-features) and add a clearly marked not-yet-implemented section pointing at ADR 0001, the trace-model plan, and the CONTEXT.md glossary. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The acronym stays; the name now leads the trace-model direction recorded in ADR 0001. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Shared references speak the Trace model: Epic/Feature/Trace/Bug kinds with legacy Task reads, spine-first sequencing in the conflict graph, depth-not-layers scope gates, trace/* branches with staged|trunk delivery modes. wtf.setup provisions the Trace type, installs the new TRACE.md template, and asks feature_scope + delivery config. Five skill dirs renamed task->trace (content migration follows). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…im them write-feature derives per-story ACs + canonical Gherkin (count set by feature_scope) and proposes a Trace Plan with Skeleton first. epic-to-features shapes its list by scope and passes both modes down. write-trace (was write-task) selects a Scenario Claim from the Feature instead of deriving Gherkin; feature-to-traces (was feature-to-tasks) validates or derives the Trace Plan and executes it in spine order, refusing layer slices. Evals migrated to the trace model. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
implement-trace drives TDD over the Scenario Claim with a Skeleton anti-slop directive and emits Revealed learnings. verify-trace runs claimed scenarios via ephemeral .feature projection with interpretive fallback and a drift gate against the Feature's canonical text. wtf.loop dispatches Traces sequentially per Feature, Features in parallel, delivery-mode aware, and re-aims after every merge through wtf.refine's new Headless mode (grow-only; shrinkage needs a human). create-pr derives bodies from story + claim and delegates PR bases to the delivery-aware branch policy. Loop/refine/create-pr evals migrated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
health gains partition-gap and missing-Skeleton signals; retro counts re-aims from refine's audit trail; changelog attributes shipped behavior via verified Scenario Claims; report-bug links failing scenarios to their claiming Trace; pr-review reviews Skeletons for lean-but-complete and treats outside-the-claim work as scope creep. README and CLAUDE.md speak the Trace model in the present tense. Own-repo issue templates refreshed (TASK.md kept as legacy); vendored shared references regenerated. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The ADR stays as the decision record; CONTEXT.md pins the vocabulary. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
xiduzo
force-pushed
the
feat/planning-mode-and-trace-model
branch
from
August 13, 2026 09:30
3ae4717 to
d199d75
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Two connected changes: planning modes for feature authoring, and the Trace model — a full replacement of the Task layer.
Planning modes (
guided/flow)skills/references/planning-mode.md:guidedasks step by step;flowderives everything it can and presents one consolidated review. Config keyplanningin.wtf/config.json, per-invocation argument override, mid-run switches both ways.wtf.write-featureandwtf.epic-to-featuresresolve the mode; flow mode fans drafting out to sub-agents and reviews once (two user gates total: confirm the list, approve the tree). All quality gates run in both modes.The Trace model (Tasks removed)
Per
docs/adr/0001-traces-replace-tasks.md— inspired by the tracer-bullet method (aihero.dev, The Pragmatic Programmer). Vocabulary pinned inCONTEXT.md.wtf.verify-traceexecutes claimed scenarios via ephemeral.featureprojection when a Gherkin runner exists, interpretive fallback otherwise.wtf.refineis the single re-aim mechanism — interactive, or headless fromwtf.loopafter every merged Trace. Headless re-aim is grow-only: it may reorder, re-batch, and add scenarios; every shrinkage needs human approval.wtf.loop.staged(trace PRs → feature branch → main) ortrunk(trace PRs → main), plusfeature_scope(single-story|grouped) — both asked once inwtf.setup, both overridable.Taskissues; write paths never create Tasks again.write-task→write-trace,feature-to-tasks→feature-to-traces,implement-task→implement-trace,verify-task→verify-trace,design-task→design-trace.All 16 shared references, 26 skills, issue templates, and 30+ eval fixtures migrated. README/CLAUDE.md speak the model in the present tense. WTF now expands to Workflow Trace Framework.
Test plan
json.loadover everyevals.json+ fixture)tasktoken is a deliberate legacy read/skill-creator test) over migrated skills/wtf.setup+/wtf.write-feature→/wtf.feature-to-traces→/wtf.loopin a consuming repo🤖 Generated with Claude Code