skills(experiment): the entry and process skills — design before data - #1061
Closed
leeovery wants to merge 4 commits into
Closed
skills(experiment): the entry and process skills — design before data#1061leeovery wants to merge 4 commits into
leeovery wants to merge 4 commits into
Conversation
The single-topic off-topic gate's ignore row named the research file whatever the session's phase. An experiment session's home for a noted concern is the live experiment's record — the variant speaks that register; the roadmap row stays discussion's alone. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Mirrors workflow-research-entry: positional args with the shared topic resolution, ensure-discovery-item at routing experiment, the triaged/completed/in-progress validation with the reconcile advisory, and a gather interview that completed research stands in for. The handoff carries identity and interview answers only — durable inputs are the processing skill's to read. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…nclude Structure and cadence mirror workflow-research-process: durable inputs read at initialisation, the register opening every session, the triage pickup beat, the presence-phase SessionEnd hooks, and a conclude flow that completes, indexes, commits, sweeps, and bridges. The walk of one experiment holds the design-before-data order — design authored, the briefing's engine-rendered approval gate freezing it, measurement advancing only after, the report executing the pre-registered rule — with the amendment protocol splitting on the first visible result and abandonment a first-class terminal. Conduct rules distilled from the design; wait releases announced in one line at conclude/abandon. The shared references the phase joins (triage pickup fold branch, off-topic reroute, seed/brief reads, closing recap, knowledge steps) learn the experiment arm, and the conventions lint's in-flight allowance comes out — the hook requirement now binds in full. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ne value per phase Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Owner
Author
|
Closed: this stack implements a superseded model of the experiment phase. The corrected design is PR #1057 ( |
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.
PR3 of the experiment-phase stack (design: #1057, stacks on #1059).
workflow-experiment-entry: phase entry mirroring research-entry — args parse,ensure-discovery-itemwith the new routing, validate/reopen/resume, carrier-aware context gather, handoff (durable inputs read at initialisation, never relayed).workflow-experiment-process: the confirmatory phase. Session opens on the register; the walk of one experiment runs conceive → design (skeleton always: question / prediction with reason / decision rule / setup; conditional controls + non-measures sections when the shape warrants) → conversational briefing → engine-rendered approval gate → freeze → run (results land as measured, deviations logged live) → report (reading separate from results, conclusion executes the pre-registered rule, exploratory labelled) → one-line verdict, or abandon with reason from any pre-terminal point. Amendment protocol: dated + re-confirmed before results, frozen for good after. Conduct reference distils the method (one primary question, instruments freeze with the design, concurrent controls, execution-first triage of disproven hypotheses, experiments measure / discussions decide). Dead-end close is epic-only, mirroring research.render off-topic-offergains--variant experiment(the ignore row named the wrong artefact).PROCESS_SKILLS_IN_FLIGHTallowance removed — check 18 binds for real.--sourcenow resolves in the prelude (the fence's{phase}/{topic}.mdshape was wrong for a series directory).Known seam, lands with PR5:
knowledge indexhas noexperimentidentity branch yet — conclude's index calls are warn-don't-block until then; the stack merges together.Gates:
npm test2666/2666 ·npm run test:cli5/5 ·npm run typecheckclean.🤖 Generated with Claude Code