Reconcile Research

Reconcile Research — Cross-Document Consistency Audit

Report-First Approval Gate

Default to report-only: present findings, evidence coverage, assumptions, recommended artifact path, and proposed file changes in a pre-approval alignment page plus a concise conversation summary for user approval before creating or updating canonical research, spec, or task files.

Do not write or overwrite synthesized deliverables until the user explicitly approves, unless the user invoked an explicit write/update/fix mode or clearly asked to write files upfront. Raw evidence capture may be persisted before analysis when reproducibility requires it; report those raw paths separately and still gate synthesized research/report writes.

When stopping for approval, build and attempt to open the alignment preview page first, then ask the user to review it and approve, question, or request adjustments. Do not include Recommended next skill, Recommended next command, or downstream routing language. The approval request itself is the next action. Only emit next-skill routing after the approved artifact has been written or updated.

Checks that research documents tell a consistent story. Finds contradictions between ICP and pricing, stale assumptions that customer feedback has invalidated, journey stages that metrics don't cover, and other cross-document gaps. Think of it as a linter for research coherence rather than structure.

Process

0. App Scope Resolution (Monorepo Support)

Before determining mode and scope, detect the app structure:

1. If $ARGUMENTS specifies an app name matching a subdirectory of research/, scope the audit to that app.
2. If research/ contains subdirectories (excluding files), list them. If the user hasn't specified a scope, reconcile per-app (each app's docs independently) plus cross-app checks.
3. If no subdirectories exist, proceed with flat structure (single-product mode).

When app scope {app} is active:

• Scan research/{app}/ for documents instead of research/
• Also check research/icp.md (cross-app overview) for cross-references
• Specs are in specs/{app}/ instead of specs/

1. Determine Mode and Scope

Parse $ARGUMENTS:

• Mode: audit (default, read-only) or fix (apply approved changes, write reconciliation report)
• Scope: all (default), icp, pricing, journey, enterprise, feedback, or specs

2. Inventory Research Documents

Scan research/ for main documents. Skip files matching *-search-log.md and *-interview.md — these are raw logs, not assertion-bearing documents.

Expected documents (not all need to exist):

• research/icp.md (cross-app overview), research/{app}/icp.md (per-app)
• research/competitive-analysis.md (or research/{app}/competitive-analysis.md)
• research/journey-map.md (or research/{app}/journey-map.md)
• research/metrics.md (or research/{app}/metrics.md)
• research/gtm.md (or research/{app}/gtm.md)
• research/monetization.md (or research/{app}/monetization.md)
• research/enterprise-icp.md (or research/{app}/enterprise-icp.md)
• research/customer-feedback.md (or research/{app}/customer-feedback.md)

When research/ contains subdirectories, scan each research/{app}/ for per-app documents.

Stop condition: If fewer than 2 research documents exist, display a message and exit — there's nothing to reconcile.

3. Extract Claims

Launch a subagent per document to extract structured claims. Each subagent reads one file and returns claims in these categories:

Each claim includes the source file, section heading, and a direct quote for traceability.

4. Cross-Reference Checks

Launch subagents per scope group to run pairwise checks. Only run checks where both documents in a pair exist.

icp scope — ICP vs all downstream (11 checks)

Requires: research/icp.md + at least one downstream document.

pricing scope — GTM vs Monetization coherence (7 checks)

Requires: research/gtm.md + research/monetization.md.

journey scope — Journey Map vs Metrics/GTM/Monetization (7 checks)

Requires: research/journey-map.md + at least one of Metrics/GTM/Monetization.

enterprise scope — Enterprise ICP vs base docs (6 checks)

Requires: research/enterprise-icp.md + at least one base document.

feedback scope — Customer Feedback propagation (5 checks)

Requires: research/customer-feedback.md + at least one other document.

Monorepo checks (3 checks)

Triggered when research/ contains app subdirectories (e.g., research/{app}/icp.md).

5. Aggregate and Classify Findings

Collect all findings from the cross-reference subagents. Classify each by severity:

6. Walk Through Findings Interactively

Present findings one at a time using the AskUserQuestion tool, ordered by severity (Errors first, then Warnings, then Info). For each finding, use AskUserQuestion with the following format:

### [Scope] — [Check Name]
**Severity**: Error | Warning | Info
**Documents**: file-a.md ↔ file-b.md

**file-a.md** says:
> [direct quote from source]

**file-b.md** says:
> [direct quote from source]

**Contradiction**: [one-sentence description of the conflict]
**Recommended resolution**: [which document to update and why, based on dependency direction]

How would you like to resolve this? (Options: accept recommendation, update the other document instead, skip/defer, or provide custom instructions)

Dependency direction informs the recommended resolution:

• Upstream contradiction (e.g., ICP vs Journey Map) → recommend updating the downstream document
• Customer feedback contradictions → recommend updating upstream (feedback is ground truth)
• Peer contradictions (e.g., GTM vs Monetization) → note both are peers and ask which is authoritative

Wait for the user's response before proceeding to the next finding. Collect all decisions into a resolution list.

7. Apply Resolutions

After walking through all findings:

• Audit mode (default): Compile a summary of all findings and user decisions. Do not modify any files. Display the summary with the user's stated resolution for each item.
• Fix mode (if fix was specified):
  1. For each existing research document that will be replaced or substantively rewritten, archive the current file first, then apply the user-approved change to the canonical research document.
  2. Write research/reconciliation-report.md as an audit trail:

# Reconciliation Report — [date]

## Resolved
- [Error/Warning description] — resolved by archiving and updating [file] per user decision: "[user's stated resolution]"

## Deferred
- [Description] — user chose to skip

## Info (no action)
- [Info items the user chose not to act on]

3. Re-run the audit to confirm fixes resolved the flagged issues.

8. Populate Next Steps

After presenting findings (audit mode) or applying resolutions (fix mode), display a ## Next Steps section with a Recommended item and Other options. Use this format:

## Next Steps

**Recommended:** [skill with most conflicts] — [N] conflicts traced to this document; re-running it will resolve the most inconsistencies at once

Other options:
- `/research-roadmap` — rebuild the ordered documentation queue after reconciliation
- `/spec-drift` — check whether code/spec drift remains after research reconciliation
- `/roadmap` — resequence implementation if reconciliation changed priorities

Output Format

Findings are presented interactively one at a time via AskUserQuestion, not as a batch report. After all findings have been walked through, display a final summary:

## Research Reconciliation — [scope] — Summary

- Documents scanned: 6
- Checks run: 28
- Findings: 2 Errors, 3 Warnings, 1 Info
- Resolved: 4, Deferred: 2

Fix mode: Same interactive walkthrough, followed by applying approved changes and writing research/reconciliation-report.md as audit trail.

Task Classification

When this skill produces follow-up work, file it by execution semantics:

• Immediately actionable implementation or documentation work goes in tasks/todo.md.
• Human-only external actions tied to automated steps go in tasks/manual-todo.md with _(blocks: Step N.X)_ or _(after: Step N.X)_; repo edits, SDK wiring, generated assets, local commands, tests, audits, and authenticated CLI/API work stay in tasks/todo.md.
• One-time condition-gated records, baselines, or future measurements go in tasks/record-todo.md with source, condition, non-blocking reason, evidence, and promotion rule.
• Cadence-based reviews, playtests, adoption checks, investor updates, retros, or docs-health checks go in tasks/recurring-todo.md with cadence, owner/agent, next due, evidence path, and escalation conditions.
• Do not put non-blocking records or recurring obligations in tasks/todo.md unless they have been explicitly promoted into current execution work.

Constraints

• Read-only by default. Only modify files when explicitly invoked with fix mode.
• Never auto-resolve contradictions. Errors always require user input on which side is correct. Use AskUserQuestion for each finding individually.
• Show evidence. Every finding must include direct quotes from both documents.
• Respect dependency direction. Upstream documents are presumed authoritative over downstream, except customer feedback which is ground truth.
• No false positives. If uncertain whether something is a real contradiction, classify it as Info, not Error.
• Skip absent documents. Only run checks where both documents in a pair exist. Never flag a missing document as an error — that's /research-roadmap's job.
• Use subagents for claim extraction (one per document) and cross-reference checks (one per scope group) to parallelize work.
• Idempotent. Running audit twice with no changes between should produce identical output.

Alignment Page

When this skill produces durable deliverables (research, specs, plans, reports, prototypes, or any document output), build a full-depth HTML alignment page at alignment/reconcile-research-{topic}.html. Use a normalized topic slug derived from the app, feature, research subject, report subject, or output filename.

Full content requirement. The alignment page must contain the complete content of every proposed markdown deliverable -- every section, every finding, every detail, every list item. It is a thorough interactive review document, not a summary. Render the full deliverable content in clean, readable HTML with appropriate hierarchy, styling, and navigation. If the skill writes multiple scoped deliverables in one run, build one alignment page that contains all deliverables with anchor-linked navigation. Durable tracker artifacts, such as research/assumption-tracker.md, remain canonical markdown outputs but must also be fully rendered into the alignment page before approval.

Dark-mode styling. Use a dark color scheme by default. Base CSS variables: --bg: #0d1117; --surface: #161b22; --border: #30363d; --text: #c9d1d9; --text-muted: #8b949e; --accent: #58a6ff; --green: #3fb950; --red: #f85149; --orange: #d29922; --purple: #bc8cff;. Apply background: var(--bg); color: var(--text); on body. Use --surface for cards, nav, and table headers. Use --border for all borders. Use --purple for question blocks and gate headings. Use --accent for links and section headings. Keep headings color: #fff or var(--accent) for hierarchy. Question block backgrounds should use #1c2333.

Alignment gates. Treat gates as explicit review sections inside the HTML page. A gate blocks finalization until its required inline questions are answered and compiled into YAML. Include every gate that applies to the skill output, and include these gate types whenever relevant: evidence coverage, assumptions/confidence, scope/non-goals, candidate/verdict decisions, artifact destination, proposed file changes, coverage checkpoint, and post-approval route.

Report-only research gates. For report-only or pre-approval research skills, the alignment page must explicitly contain evidence coverage, assumptions/confidence, recommended path, proposed file changes, and approval gates before any canonical research, spec, or task file is created or updated.

Required inline questions. Each gate must contain at least one required inline question placed directly under the content it governs, inside a visually distinct question block. Each question must use radio-button inputs and include two standing options after the skill-generated choices: "Other / None of the above" backed by a multi-line text box for free-form input, and "Need clarification" backed by an optional notes box where the user can explain what is unclear. When any radio option other than "Other" or "Need clarification" is selected, show an optional "Additional notes" text box beneath it so the user can qualify their choice. Generate questions based on what genuinely needs user input -- do not add filler questions. Do not create a separate bottom "Decisions & Clarifications" section.

Gate YAML contract. At the bottom of the page, include a "Compile Answers" button that aggregates answers from all inline gate questions throughout the page, including free-text notes. The button remains disabled until every required question has a selection, shows a count of remaining unanswered questions, and scrolls to the first unanswered question if clicked early. When every question is answered, generate a structured YAML block with one item per gate answer using this stable shape: section, gate_type, status (answered, other, or needs-clarification), answer, optional notes, and optional target_artifact or target_path when the gate controls file output. After successful compilation, automatically attempt to copy the YAML to the clipboard with the Clipboard API, display copy status, and display the YAML in a read-only textarea with an explicit "Copy YAML" button. The copy button must retry clipboard copy when supported and fall back to selecting the textarea contents when clipboard access is unavailable or blocked.

Pre-approval stop. Before user approval, the next action is review of the HTML alignment page, not downstream routing. Ask the user to review the page and provide the compiled YAML answers. Do not include Recommended next skill, Recommended next command, or downstream routing language until after compiled YAML has been provided and the approved artifacts have been written or updated.

Diff highlighting on updates. When the agent updates an existing alignment page after receiving compiled answers, highlight what changed since the previous version. The agent chooses inline annotation or side-by-side layout per situation.

Archiving. Before replacing an existing alignment page, archive it to docs/history/archive/YYYY-MM-DD/HHMMSS/alignment/reconcile-research-{topic}.html.

Browser open. Attempt to open the resulting HTML page in the browser and report whether the open succeeded or was blocked. A blocked browser-open attempt does not make the skill fail when the files were written correctly.

Archive-First Replacement Policy

• Before replacing or substantively rewriting an existing canonical research/spec document (research/**/*.md, specs/**/*.md, or docs/specifications/**/*.md), copy the current file to docs/history/archive/YYYY-MM-DD/HHMMSS/<original-relative-path>.
• Preserve the archived snapshot exactly as it existed before the change; do not edit the archived copy after creating it.
• After the archive snapshot exists, write the updated document to the original canonical path.
• Report both the archive path and the updated canonical path in the final output.
• New files do not need archive snapshots. Append-only updates do not need archive snapshots unless an existing section is regenerated or rewritten.
• Keep any existing user approval requirement before overwriting or replacing a document; archiving does not replace asking when the skill already requires approval.

Default Shipping Contract

• Default next-step routing: when reporting completion, include either Recommended next skill: <command> or the two-line pair **Next work:** <specific task or "none"> and **Recommended next command:** <one command or route> so the next operator has a concrete handoff.
• If this skill creates or modifies tracked repository files, finish by committing and pushing all intended changes to the repository primary branch (main when present, otherwise master) before stopping, even if the user did not explicitly ask for commit/push.
• Do not leave tracked changes or unpushed commits behind. If unrelated tracked work is already present, either include it in sensible commits too or stop and explain the blocker.
• This contract does not override stricter safety rules about secrets, destructive history changes, release publication/tag confirmation, or production deploy confirmation.