feat(agent-skill): v1.3 PRD Elicitation Enhancement#4
Merged
Conversation
Transform prd-guide pack from teaching ODD to actively eliciting PRDs through structured questioning. New in v1.3 INSTRUCTIONS.md: - Agent Role Declaration (You extract. You do not invent.) - PRD Stage Typing table (6 types with evidence/ambiguity expectations) - Asset Intake Contract (4 asset types with guidance) - 8-phase Interview Loop (was 7 stages): - Phase 0: Stage Identification (NEW) - Phase 1: Orient (NEW) - Phase 2: Inventory (NEW) - Phase 6: Ambiguity Capture (NEW) - Resequenced flow: Inventory before Outcome - Updated example dialogue demonstrating new elicitation flow Canon sources unchanged (v0.8.0, same hashes as v1.2.4).
Deploying klappy-dev-website with
|
| Latest commit: |
2d5f015
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://9e2f94f6.klappy-dev-website.pages.dev |
| Branch Preview URL: | https://agent-skill-v1-3-attempt-001.klappy-dev-website.pages.dev |
Deploying klappy-dev-agent-skill with
|
| Latest commit: |
2d5f015
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://65007ecc.klappy-dev-agent-skill.pages.dev |
| Branch Preview URL: | https://agent-skill-v1-3-attempt-001.klappy-dev-agent-skill.pages.dev |
added 2 commits
January 21, 2026 15:23
HTTP 200 verified on preview URLs. All Definition of Done criteria met.
- Update current champion to v1.3 - Mark v1.2.4 as superseded - Add H0008 history entry for v1.3 champion
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.
|
|
||
| **ODD principle**: Uncertainty acknowledged early is cheaper than uncertainty discovered late. | ||
|
|
||
| --- |
There was a problem hiding this comment.
Interview loop missing non-goals gathering phase
Medium Severity
The 8-phase Interview Loop (Phases 0-6) lacks explicit questions for gathering non-goals, yet the Phase 7 PRD template includes a Non-Goals (Out of Scope) section, and the success criteria require "Honest scope — Non-goals prevent scope creep." The v1.2.4 version had dedicated "Stage 3: Non-Goals and Scope" with specific probing questions like "What related features might someone assume are included but aren't?" This stage was marked as "Consolidated" into Phase 7, but Phase 7 only assembles the draft without gathering information. An agent following the phases strictly will never ask about scope exclusions.
Additional Locations (2)
klappy
added a commit
that referenced
this pull request
Apr 19, 2026
Live validation of telemetry_policy canary (klappy/oddkit#106) against prod surfaced three gaps the original contract didn't name explicitly enough: 1. Response envelope shape is part of the contract. A tool that returns {action, result} but omits server_time/assistant_text/debug breaks the time-discipline system even if governance_source is present. Added as Runtime Invariant #3. 2. canon_url parameter must be in the Zod schema, not just documented as a concept. MCP silently strips unknown parameters. The canary shipped with schema={} and canon_url was unreachable. Added as Runtime Invariant #4. 3. Live-smoke against the MCP endpoint is a ship-blocker, not a nice-to-have. Internal parser tests passed while the tool shipped with broken envelope and silent param stripping. Added as Runtime Invariant #7, template referenced. Refactor Implications section expanded to a 7-point checklist and acknowledges the canary's partial completion + follow-up PR as the first documented test of the contract. Follow-up PR that closes the canary gaps: klappy/oddkit#108.
klappy
added a commit
that referenced
this pull request
Apr 19, 2026
Adds validation as a first-class epistemic mode, peer to exploration, planning, and execution. Prior canon described validation extensively (oddkit_validate tool, qa-validation case study, many process refs) but never named it as a distinct mode with its own truth conditions and non-collapse obligations. The three-mode framing implicitly treated validation as a phase of execution, which produced the 'validating mid-build' failure pattern — noticing concerns during execution and surfacing them as inline pivots instead of carrying them to a dedicated post-execution review. This is the second half of the mode-collapse pattern documented in canon/constraints/mode-discipline-and-bottleneck-respect. The first half is planning-into-execution (inline clarifying questions). The second half is validation-into-execution (inline concerns). Both externalize cost to the operator's attention. Both feel like care to the agent performing them. Both are the same throughput violation. New canon: - canon/definitions/validation-as-epistemic-mode.md (tier:1) Full contract for validation as a fourth mode. Names purpose, characteristics, truth condition, obligations, primary risk, valid/invalid moves. Defines the exec→val→(accept|iterate|pivot) rhythm. Extends the non-collapse rule to cover all six pairings. Revisions: - canon/definitions/epistemic-modes.md — blockquote and section heading updated from three to four modes; validation added as mode #4; non-collapse rule extended to name validation-specific collapses. - canon/constraints/mode-discipline-and-bottleneck-respect.md — 'Three Modes' section becomes 'Four Modes'; non-collapse contract extended with 'execution pretending to validate', 'validation pretending to plan', 'validation pretending to execute'; the rhythm named explicitly. - canon/bootstrap/model-operating-contract.md — summary and Mode Discipline section updated to four modes; validation-mode invalid moves listed; exec→val→iter rhythm named. - docs/appendices/mode-separated-conversations.md — adds Validation Conversations section mirroring Planning and Execution sections. - docs/examples/project-instructions-template.md — public template updated to four modes with validation-mode invalid moves list. Sequencing: this PR layers on top of klappy.dev#104 (merged), which introduced the mode-discipline canon and model-operating-contract bootstrap in their original three-mode framing. Landing validation as a separate PR keeps the delta reviewable and preserves the commit history of the original #104 work. Origin: this work came out of a post-canary review session. The gauntlet-then-fix rhythm (execution produces → validation reviews → iteration pivots) worked cleanly in practice while the canon had no name for the middle step. Naming validation as its own mode closes the gap between how the system actually works and how it is documented.
This was referenced Apr 29, 2026
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
Transform the prd-guide pack from an informational resource into an interrogative system:
What is New in v1.3
Test Plan
Note
Introduces the v1.3 PRD Guide pack and promotes it to champion, shifting from guidance to an elicitation workflow.
INSTRUCTIONS.mdfor v1.3 with an 8-phase interview loop, explicit agent role (“You extract. You do not invent.”), PRD stage typing, asset intake contract, and ambiguity capture; resequences flow (Inventory → Constraints → Outcome)public/agent-skill/v1.3/prd-guide-pack.mdand updatespublic/agent-skill/latest/(pack + README, metadata)H0008-v1.3-champion.mdATTEMPT.md, provenance-packedprd-guide-pack.md, compile logs, instructions diff, deployment verification with HTTP 200)Written by Cursor Bugbot for commit 2d5f015. This will update automatically on new commits. Configure here.