Rishikesh Ranjan — rishikeshranjan.com

// 9-phase lifecycle

Fix

/ttm-fix

Fix phase: root cause analysis, fix brief, re-produce in isolated context, re-verify. Capped at 3 attempts per asset. Use when assets fail review.

phase 5manual invocation
allowed-tools:ReadWriteBashGlobGrepTaskAskUserQuestion
SKILL.md ↗

Overview

ttm-fix is the surgical repair phase of the takeToMarket lifecycle. When assets fail review, it reads the verification findings, opens the offending asset, performs root cause analysis, generates a targeted fix brief, re-produces the asset in an isolated Task() context, and re-runs all 10 verification gates. It does not regenerate from scratch — it patches.

Invoke it with ttm-fix [campaign-slug]. The fix-verify loop is capped at 3 attempts per asset. On a successful fix, the asset is auto-approved to ship-ready; if the cap is hit without resolution, the asset is escalated to needs-human-fix with a failure-pattern analysis and suggested manual edits.

Why it matters: regenerating an asset for a small drift wastes context and loses good prose that was almost right. Fix is the surgical tool; produce is the hammer. Most assets need one or two fix passes before shipping, and that loop is where positioning drift gets caught before it leaves the repo.

How it works

The skill reads and follows the workflow at ${CLAUDE_PLUGIN_ROOT}/workflows/lifecycle/fix.md. It performs root cause analysis on assets marked needs-fix during review, generates targeted fix briefs, re-produces in isolated Task() context, re-verifies against all 10 gates (D-06), shows results at each iteration (D-07), and caps at 3 attempts with escalation (D-08, LIFE-12). It auto-approves to ship-ready on a successful fix (D-14).

Step 0: First-run inline education

Read .taketomarket/CONFIG.md. Parse first_run_seen (object) and inline_education (boolean, default true).

If inline_education is false: skip this step. Else if first_run_seen.ttm-fix is not true, print the explainer below verbatim, then mark this skill as seen:

node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" first-run mark ttm-fix

Use this exact check (bash) to decide whether to print: node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" first-run check ttm-fix --raw — the JSON seen field is true once the explainer has run before.

Explainer for /ttm-fix

/ttm-fix reads the verification findings, opens the offending asset, applies targeted corrections, and re-runs only the affected gates. It's deliberately narrow — it does not regenerate from scratch, it patches. The fix-verify loop continues until either all gates pass or you escalate.

Why it matters: regenerating an asset for a small drift wastes context and loses good prose that was almost right. Fix is the surgical tool; produce is the hammer. Most assets need one or two fix passes before shipping, and that loop is where positioning drift gets caught before it leaves the repo.

(Canonical source: references/inline-education-blurbs.md. Embedded verbatim because workflows do not @-resolve files at runtime.)

Required reading for the workflow: ${CLAUDE_PLUGIN_ROOT}/references/context-loading.md, ${CLAUDE_PLUGIN_ROOT}/gates/base-gates.md, ${CLAUDE_PLUGIN_ROOT}/gates/gate-evaluation.md, and ${CLAUDE_PLUGIN_ROOT}/agents/ttm-producer.md.

Text-Mode Detection

Text mode (--text flag): Set TEXT_MODE=true if --text is present in $ARGUMENTS or if the runtime is not Claude Code. When TEXT_MODE is active, replace every AskUserQuestion call with a plain-text numbered list.

Detection:

if echo "$ARGUMENTS" | grep -q -- '--text'; then TEXT_MODE=true; fi

If AskUserQuestion tool is not available in the current runtime, set TEXT_MODE=true. When TEXT_MODE is active, replace each AskUserQuestion with a plain-text numbered list:

[HEADER]
[QUESTION]
  1. [OPTION_1_LABEL] -- [OPTION_1_DESCRIPTION]
  2. [OPTION_2_LABEL] -- [OPTION_2_DESCRIPTION]
  ...
Type the number of your choice:

Step 1: Load Context

takeToMarket > LOADING CONTEXT

Extract SLUG from $ARGUMENTS (strip --text flag if present):

SLUG=$(echo "$ARGUMENTS" | sed 's/--text//g' | xargs)

If SLUG is empty, error: "Usage: /ttm-fix [campaign-slug]. Provide a campaign slug." Exit.

Load Tier 1 summaries from all 9 reference files (lines 1 to <!-- END_SUMMARY -->):

  • .taketomarket/POSITIONING.md
  • .taketomarket/BRAND.md
  • .taketomarket/ICP.md
  • .taketomarket/CHANNELS.md
  • .taketomarket/STATE.md (frontmatter only)
  • .taketomarket/CALENDAR.md
  • .taketomarket/COMPETITORS.md
  • .taketomarket/METRICS.md
  • .taketomarket/LEARNINGS.md

Load Tier 2 (full content) for gate evaluation (same as verify per context-loading.md):

  • .taketomarket/POSITIONING.md (needed for GATE-01 Positioning Drift)
  • .taketomarket/BRAND.md (needed for GATE-02 Claim Accuracy, GATE-03 Voice Drift)
  • .taketomarket/ICP.md (needed for GATE-09 ICP Fit)
  • .taketomarket/COMPETITORS.md (needed for GATE-08 Competitor Collision)
  • .taketomarket/CHANNELS.md (needed for GATE-06 UTM Hygiene)

Load campaign-specific files (always full-load per context-loading.md rule 4):

  • .taketomarket/CAMPAIGNS/${SLUG}/STATE.md
  • .taketomarket/CAMPAIGNS/${SLUG}/BRIEF.md

Load MANIFEST.json: Read .taketomarket/CAMPAIGNS/${SLUG}/MANIFEST.json. If the file does not exist, error: "No production manifest found for campaign '${SLUG}'. Run /ttm-produce first." Exit.

Load VERIFICATION.md: Read .taketomarket/CAMPAIGNS/${SLUG}/VERIFICATION.md. If the file does not exist, error: "No verification report found for campaign '${SLUG}'. Run /ttm-verify first." Exit.

Load LEARNINGS.md Tier 2 (for root-cause taxonomy categories): .taketomarket/LEARNINGS.md (full content — needed for root-cause category matching).

Step 2: Validate Campaign State

takeToMarket > VALIDATING CAMPAIGN

Check campaign exists:

node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign state "${SLUG}" --raw

If result shows exists: false: Tell the user the campaign does not exist and suggest running /ttm-new-campaign first. Exit.

Read campaign state. Check the phase field. If phase is NOT "reviewed": Warn the user — "Campaign is in phase '${PHASE}'. Expected 'reviewed' before fix. Run /ttm-review first." Exit.

Determine fix run number: Read fix.run_count from state. If null or 0, set RUN_NUMBER=1. Otherwise set RUN_NUMBER to fix.run_count + 1.

Step 3: Identify Assets Needing Fix

takeToMarket > IDENTIFYING ASSETS

Parse MANIFEST.json for assets where review_status == "needs-fix".

Collect from both hero and derivatives entries. For each matching asset, record:

  • asset_id, name, file, type, channel, playbook
  • Load its review feedback file: .taketomarket/CAMPAIGNS/${SLUG}/REVIEW-FEEDBACK-${NAME}.md. If the feedback file does not exist, log warning and use VERIFICATION.md failures only.

If no assets need fix: "No assets need fixing. All assets are either approved or rejected. Run /ttm-ship to proceed." Exit.

Display summary:

takeToMarket > FIX LOOP

Assets needing fix: ${COUNT}
- [ASSET_NAME_1] (severity from review feedback)
- [ASSET_NAME_2] (severity from review feedback)

Step 4: Initialize Fix Log

Check if FIX-LOG.md exists in campaign directory: .taketomarket/CAMPAIGNS/${SLUG}/FIX-LOG.md.

If not, create it from ${CLAUDE_PLUGIN_ROOT}/templates/fix-log.md template:

  • Replace [SLUG] with campaign slug
  • Replace [ISO_TIMESTAMP] with current timestamp

If FIX-LOG.md already exists, read it to determine prior attempt counts per asset. Parse existing entries to find the highest attempt number for each asset.

Step 5: Fix Loop Per Asset

For each asset with review_status == "needs-fix":

Initialize attempt_count from prior attempts in FIX-LOG.md (0 if no prior attempts).

WHILE attempt_count < 3 AND asset_status == "needs-fix": Increment attempt_count.

5a. Root Cause Analysis (D-05)

Load: review feedback file + VERIFICATION.md failures for this asset + BRIEF.md. Analyze against LEARNINGS.md root-cause taxonomy. Propose one of these 7 categories:

  • positioning-drift — Asset contradicts or extends beyond POSITIONING.md
  • weak-hook — Opening/hook does not grab ICP attention
  • wrong-channel — Content format mismatched for target channel
  • bad-timing — References or angles are dated or poorly timed
  • unverifiable-claim — Contains claims not in approved proof points
  • broken-funnel — CTA, conversion path, or next-step is missing/broken
  • creative-fatigue — Content is generic, template-like, or lacks originality

Present to user via AskUserQuestion (or text-mode):

Proposed root cause for [ASSET_NAME]: [CATEGORY]
Explanation: [Why this root cause fits based on review feedback and gate failures]

1. Confirm -- proceed with this root cause
2. Correct -- select a different root cause

If user selects "Correct", present numbered list of all 7 categories and let them choose.

5b. Generate Fix Brief

Read template from ${CLAUDE_PLUGIN_ROOT}/templates/fix-brief.md. Fill all placeholders:

  • [SLUG] → campaign slug
  • [ASSET_NAME] → asset name from manifest
  • [ATTEMPT_NUMBER] → current attempt (1, 2, or 3)
  • [ROOT_CAUSE_CATEGORY] → confirmed root cause category
  • [ROOT_CAUSE_EXPLANATION] → explanation text
  • [BRIEF_PATH] → .taketomarket/CAMPAIGNS/${SLUG}/BRIEF.md
  • [FAILURE_LIST] → extract all WARN/FAIL findings for this asset from VERIFICATION.md, plus review feedback items from REVIEW-FEEDBACK-${NAME}.md
  • [PASSING_LIST] → extract all PASS findings for this asset from VERIFICATION.md (per RESEARCH Pitfall 3: these are preservation constraints — the producer MUST NOT break these)
  • [CORRECTIONS_LIST] → derive specific corrections from root cause + review feedback
  • [ISO_TIMESTAMP] → current timestamp

Write fix brief to: .taketomarket/CAMPAIGNS/${SLUG}/FIX-BRIEF-${ASSET_ID}-attempt-${N}.md.

5c. Re-Produce Asset (Task() subagent)

Read agent prompt template from ${CLAUDE_PLUGIN_ROOT}/agents/ttm-producer.md. Fill placeholders — CRITICAL: use the FIX BRIEF path as [BRIEF_PATH], NOT the original BRIEF.md:

  • [BRIEF_PATH] → .taketomarket/CAMPAIGNS/${SLUG}/FIX-BRIEF-${ASSET_ID}-attempt-${N}.md
  • [POSITIONING_PATH] → .taketomarket/POSITIONING.md
  • [BRAND_PATH] → .taketomarket/BRAND.md
  • [ICP_PATH] → .taketomarket/ICP.md
  • [PLAYBOOK_PATH] → playbook path from MANIFEST.json or "none"
  • [OUTPUT_PATH] → .taketomarket/CAMPAIGNS/${SLUG}/${ASSET_FILE} (overwrite failing version)
  • [ASSET_TYPE] → from MANIFEST.json
  • [CHANNEL] → from MANIFEST.json
  • [HERO_PATH] → hero asset path if this is a derivative, else "none"

Call Task() with populated prompt. Wait for completion. Verify the re-produced file exists and has content:

test -s ".taketomarket/CAMPAIGNS/${SLUG}/${ASSET_FILE}"

If file missing/empty: log as failed attempt in FIX-LOG.md, continue loop.

5d. Re-Verify Against All 10 Gates (D-06)

Follow the gate evaluation pattern from verify.md Step 4. For THIS SINGLE ASSET only, for each of the 10 gates (in order from base-gates.md):

  1. Load the gate-specific reference data (same as verify.md)
  2. Evaluate per gate-evaluation.md instructions
  3. Record structured output: gate, tier, result, findings[]

Aggregate results for this asset: PASS/WARN/FAIL per gate.

5e. Present Result to User (D-07)

Capture the "before" results from the original VERIFICATION.md for this asset. Display gate summary table for this attempt:

takeToMarket > FIX ATTEMPT ${N}/3: [ASSET_NAME]

Root cause: [CATEGORY]

| Gate | Before | After |
|------|--------|-------|
| Positioning Drift (GATE-01) | [BEFORE] | [AFTER] |
| Claim Accuracy (GATE-02) | [BEFORE] | [AFTER] |
| Voice Drift (GATE-03) | [BEFORE] | [AFTER] |
| Outcome Alignment (GATE-04) | [BEFORE] | [AFTER] |
| Funnel Integrity (GATE-05) | [BEFORE] | [AFTER] |
| UTM Hygiene (GATE-06) | [BEFORE] | [AFTER] |
| Compliance (GATE-07) | [BEFORE] | [AFTER] |
| Competitor Collision (GATE-08) | [BEFORE] | [AFTER] |
| ICP Fit (GATE-09) | [BEFORE] | [AFTER] |
| Format Correctness (GATE-10) | [BEFORE] | [AFTER] |

Determine fix result:

If ALL gates PASS (or only Tier 2 WARNs with no FAILs): Fix successful.

Fix successful! Auto-approving [ASSET_NAME] to ship-ready. (D-14)

Set asset_status = "ship-ready". Break loop.

If any Tier 1 gates FAIL: Fix attempt failed. If attempt_count < 3, present to user via AskUserQuestion (or text-mode):

Attempt ${N}/3 still has failures:
[List of remaining FAIL findings]

1. Continue fixing -- try again with adjusted approach
2. Approve anyway -- accept current state and mark ship-ready
3. Adjust feedback -- provide new specific feedback for next attempt
  • "Continue fixing": loop continues with next attempt
  • "Approve anyway": set asset_status = "ship-ready", break loop
  • "Adjust feedback": collect new freeform feedback via AskUserQuestion, append to review feedback file, loop continues with next attempt

5f. Log Attempt to FIX-LOG.md

Append to .taketomarket/CAMPAIGNS/${SLUG}/FIX-LOG.md:

## Asset: [ASSET_NAME]

### Attempt [N]
- **Date:** [ISO_TIMESTAMP]
- **Root cause:** [CATEGORY] -- [EXPLANATION]
- **Fix brief:** FIX-BRIEF-[ASSET_ID]-attempt-[N].md
- **Gate results after fix:**
  | Gate | Before | After |
  |------|--------|-------|
  | [GATE_NAME] | [BEFORE] | [AFTER] |
  | ... |
- **Outcome:** [passed | failed -- reason | approved-by-user]

END WHILE

Step 6: Handle 3-Attempt Escalation (D-08, LIFE-12)

If attempt_count == 3 AND asset still has gate failures (asset_status still "needs-fix"):

takeToMarket > ESCALATION: [ASSET_NAME]

3 fix attempts exhausted. Presenting attempt history:

Read FIX-LOG.md entries for this asset. Display all 3 attempts:

### Attempt 1
Root cause: [CATEGORY]
Gate results: [summary -- N PASS / N WARN / N FAIL]
Fix brief: FIX-BRIEF-[ASSET_ID]-attempt-1.md

### Attempt 2
Root cause: [CATEGORY]
Gate results: [summary]
Fix brief: FIX-BRIEF-[ASSET_ID]-attempt-2.md

### Attempt 3
Root cause: [CATEGORY]
Gate results: [summary]
Fix brief: FIX-BRIEF-[ASSET_ID]-attempt-3.md

### Failure Pattern Analysis
[AI analysis of what consistently failed across all 3 attempts -- identify the
recurring gates, whether the same root cause persists, and any oscillation patterns]

### Suggested Manual Edits
Based on the pattern of failures:
1. [Specific edit suggestion with file location and what to change]
2. [Specific edit suggestion]

Asset status: needs-human-fix
You can manually edit the file and re-run /ttm-verify ${SLUG}

Set asset review_status to "needs-human-fix" in memory.

Step 7: Update MANIFEST.json

Read MANIFEST.json from .taketomarket/CAMPAIGNS/${SLUG}/MANIFEST.json. Update review_status for each processed asset:

  • "ship-ready" if fix succeeded or user approved anyway
  • "needs-human-fix" if 3-attempt cap reached without resolution

Also add fix_attempts count per asset:

{
  "hero": {
    ...existing fields...,
    "review_status": "ship-ready|needs-human-fix",
    "fix_attempts": 2
  },
  "derivatives": [
    {
      ...existing fields...,
      "review_status": "ship-ready|needs-human-fix",
      "fix_attempts": 3
    }
  ]
}

Write updated MANIFEST.json back to disk.

Step 8: Update VERIFICATION.md

After the final re-verification of each asset, update VERIFICATION.md with the latest gate results. Overwrite the file with the new results (same pattern as verify.md Step 8).

Use the templates/verification-report.md format. Include results from the most recent re-verification run for fixed assets. Assets not processed by fix retain their original verification results.

Step 9: Update Campaign State

TIMESTAMP=$(node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" timestamp --raw)
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" fix.run_count "${RUN_NUMBER}"
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" fix.last_run "$TIMESTAMP"
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" fix.overall_result "[all-fixed|partial|escalated]"
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" phase fixed
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" phase.fixed "$TIMESTAMP"

fix.overall_result logic:

  • all-fixed — all needs-fix assets now ship-ready (including user "approve anyway")
  • partial — some fixed, some still need human fix
  • escalated — all needs-fix assets hit the 3-attempt cap

Also update gate fields in STATE.md with latest verification results for fixed assets (same pattern as verify.md Step 9):

node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.positioning_drift [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.claim_accuracy [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.voice_drift [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.outcome_alignment [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.funnel_integrity [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.utm_hygiene [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.compliance [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.competitor_collision [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.icp_fit [pass|warn|fail]
node "${CLAUDE_PLUGIN_ROOT}/bin/ttm-tools.cjs" campaign update "${SLUG}" gate.format_correctness [pass|warn|fail]

For each gate, use the worst result across all assets (including both fixed and non-fixed).

Step 10: Display Completion

takeToMarket > FIX COMPLETE

Run: ${RUN_NUMBER}
Assets processed: ${TOTAL}
  Fixed (ship-ready): ${FIXED_COUNT}
  Needs human fix: ${ESCALATED_COUNT}

Fix log: .taketomarket/CAMPAIGNS/${SLUG}/FIX-LOG.md

If any ship-ready assets:

Next: Run /ttm-ship ${SLUG} to launch approved assets

If any needs-human-fix assets:

Edit the flagged files manually, then run /ttm-verify ${SLUG} to re-check

Success criteria

  • All needs-fix assets from MANIFEST.json processed through fix loop
  • Root cause proposed from 7-category taxonomy with user confirmation (D-05)
  • Fix brief generated per attempt with failure list, preservation constraints, and corrections
  • Task() re-production uses fix brief path (NOT original brief) as [BRIEF_PATH]
  • All 10 gates re-run after each fix per gate-evaluation.md (D-06)
  • Results shown to user per iteration with before/after comparison (D-07)
  • Auto-approve to ship-ready on all gates passing (D-14)
  • 3-attempt cap enforced with escalation display (D-08, LIFE-12)
  • FIX-LOG.md records every attempt with root cause, fix brief, and gate results
  • MANIFEST.json updated with review_status and fix_attempts per asset
  • Campaign STATE.md updated with fix.run_count, fix.last_run, fix.overall_result
  • VERIFICATION.md updated with latest gate results after fix

Outputs

  • .taketomarket/CAMPAIGNS/${SLUG}/FIX-LOG.md (fix attempt history — append-only)
  • .taketomarket/CAMPAIGNS/${SLUG}/FIX-BRIEF-*-attempt-*.md (fix briefs per attempt — persistent)
  • .taketomarket/CAMPAIGNS/${SLUG}/MANIFEST.json (updated with fix results per asset)
  • .taketomarket/CAMPAIGNS/${SLUG}/VERIFICATION.md (updated with latest gate results)

What if this doesn't fit?

Looks like /ttm-fix can't do that yet.

  • Want a new skill? /ttm-request-skill
  • Existing skill needs work? /ttm-improve-skill