Fix index mode on the four single-file plugin payloads

brief-schema.md and eco-floors.yaml were still recorded as mode 120000 in both
plugins. Replacing a symlink's content on a core.symlinks=false checkout swaps
the bytes but leaves the index mode alone, so git was carrying a symlink whose
target was the entire file body. A Linux checkout would have tried to create
that link and produced a broken path where the ECO floor registry should be --
reintroducing the "ECO floor registry not found" failure this change set out to
fix.

Forced to 100644 via update-index --cacheinfo. No symlinks remain under
plugins/.

Author: cgallic

plugins/kai-marketing-os-v2/harness/brief-schema.md

@@ -0,0 +1,154 @@
+# Content Brief Schema
+
+Every content task requires a completed brief. No brief = no write.
+Save to `/tmp/harness_brief.json` before passing to write agent.
+
+---
+
+## Required Fields
+
+```json
+{
+  "target_site": "kaicalls | buildwithkai | abp | meetkai | connorgallic | vocalscribe",
+  "target_keyword": "primary keyword exactly as it appears in title",
+  "secondary_keywords": ["2-3 supporting terms"],
+  "format": "blog | linkedin | email | tiktok | ad | press",
+  "persona": "archetype name from knowledge/personas/",
+  "persona_evidence_status": "evidence_backed | directional | hypothesis",
+  "persona_evidence_sources": [
+    {
+      "source_type": "interview | sales_call | support_ticket | review | survey | analytics | third_party_research | hypothesis",
+      "source": "URL, file path, CRM note, or research artifact",
+      "retrieved_or_observed_date": "YYYY-MM-DD",
+      "confidence": "high | medium | low"
+    }
+  ],
+
+  "current_rank": "position for target keyword, or not ranking",
+  "monthly_impressions": 0,
+  "current_ctr": 0.0,
+  "data_provenance_mode": "sales_external | onboarding_connected | internal_demo | not_applicable",
+  "quantitative_claims": [
+    {
+      "claim": "exact claim that may appear in output",
+      "source": "URL, file path, collector output, or internal measurement",
+      "retrieved_or_observed_date": "YYYY-MM-DD",
+      "evidence_tier": "official_requirement | official_best_practice | law_regulation_court_status | academic_study | vendor_platform_study | practitioner_benchmark | internal_measurement | inference_hypothesis | missing_data",
+      "confidence": "high | medium | low"
+    }
+  ],
+  "data_gaps": ["known missing data that must not be guessed"],
+  "competitor_url": "top-ranking URL we're competing against",
+  "competitor_weakness": "specific gap in their content — not vague",
+
+  "angle": "specific frame — not AI for law firms, but why law firms lose 40% of leads after 5pm",
+  "hook_options": [
+    "Hook variant 1",
+    "Hook variant 2",
+    "Hook variant 3"
+  ],
+  "audience_pain": "the single biggest frustration of this persona",
+  "proof_available": "data, stories, or examples we can use",
+  "proof_inventory": [
+    {
+      "proof_type": "customer_quote | product_data | case_study | demo | review | benchmark | legal_requirement | platform_requirement",
+      "source": "URL, file path, collector output, or note",
+      "claim_allowed": "what this proof can safely support"
+    }
+  ],
+  "cta": "what we want them to do after reading",
+
+  "ad_concept_bench": {
+    "use_when": "required for paid ad batches; omit for non-ad content",
+    "personas": ["3-5 situational or psychographic personas"],
+    "desires": ["3-5 customer-language desired outcomes"],
+    "angles": ["4-8 story, proof, or mechanism frames"],
+    "concept_math": "personas x desires x angles = total possible concepts",
+    "portfolio_rule": "60% winners, 30% adjacent tests, 10% experiments",
+    "selected_concepts": [
+      {
+        "concept_id": "PDA-persona-desire-angle-01",
+        "persona": "who this ad speaks to",
+        "desire": "the progress they want",
+        "angle": "the story/proof/frame",
+        "awareness_stage": "unaware | problem-aware | solution-aware | product-aware | most-aware",
+        "format": "video | static | carousel | UGC | founder | demo",
+        "hook": "first line or first 3 seconds",
+        "portfolio_bucket": "winner | adjacent | experiment",
+        "kill_rule": "named threshold before launch"
+      }
+    ]
+  },
+
+  "word_count_target": 1400,
+  "publish_date": "YYYY-MM-DD",
+  "internal_links": [
+    "https://site.com/existing-post-1",
+    "https://site.com/existing-post-2"
+  ]
+}
+```
+
+---
+
+## Word Count Targets by Format
+
+| Format | Target |
+|--------|--------|
+| Blog post | 1200–1800 |
+| LinkedIn article | 700–1000 |
+| Email | 300–500 |
+| TikTok script | 150–300 (spoken words) |
+| Meta ad | 50–150 |
+| Press release | 400–600 |
+
+---
+
+## Persona Reference
+
+Load the matching file from `knowledge/personas/` before completing the brief.
+Available archetypes: check `ls knowledge/personas/` for current list.
+
+Personas must be labeled:
+
+- `evidence_backed`: supported by interviews, sales calls, support tickets, analytics, reviews, or credible research.
+- `directional`: supported by partial evidence but still needs validation.
+- `hypothesis`: useful for creative exploration only; not enough for client-facing strategy claims.
+
+Do not present a persona pain, budget authority, buying trigger, or objection as fact unless the evidence source is listed.
+
+---
+
+## Outcome Declaration (ECO O1 — required before writing)
+
+A baseline recorded after the piece ships is not a baseline, and a threshold chosen after seeing the result is not a threshold. Capture both here, before the first draft.
+
+```yaml
+outcome:
+  metric: organic_clicks           # what this piece is supposed to move
+  source: google_search_console    # the authoritative system, not a dashboard screenshot
+  baseline: 0                      # the pre-state, measured now
+  threshold: 120                   # the number that counts as success
+  window_days: 30                  # when it gets read
+  owner: Connor                    # who reads it
+  attribution: observational       # observational | control | holdout | geo_split | switchback
+```
+
+This block becomes the `outcome_baseline` evidence entry on the piece's ECO record. Without it the work can reach SHIPPED but never CLOSED.
+
+Floors per format: `harness/eco-floors.yaml` · Doctrine: `docs/system/eco-completion-standard.md`
+
+---
+
+## Brief Validation
+
+Before write agent starts, validate:
+- `outcome` block is complete (metric, source, baseline, threshold, window, owner) and recorded before drafting
+- All required fields present and non-empty
+- `hook_options` has exactly 3 variants
+- `competitor_weakness` is specific (≥20 words), not generic
+- `angle` is differentiated from `target_keyword` (not just a restatement)
+- `proof_available` references actual data or a named example
+- `persona_evidence_status` is present and matches the sources
+- Every quantitative claim has a source, evidence tier, date, and confidence label
+- Missing data is listed in `data_gaps`, not replaced with a benchmark