Evaluate content

Content Eval — Orchestrator

<!-- BUDGET_EXCEPTION: Eval skills carry artifact-schema-as-contract (8 body sections + 8-col results row + cross-stack consumer contract) that is load-bearing and cannot move to references/. Content-eval also surfaces primary-platform scoping + engagement-quality discrimination rules (which are the point of an organic-content eval). Cycle ledger discipline requires the schema be visible in the SKILL.md body. ~800 tokens over the standard cap is the legitimate cost (matches evaluate-campaign sibling). -->

Evaluation skill. Converts published organic-content evidence into a cycle snapshot + ledger row + narrowly-scoped next action inside an existing eval loop. One primary platform per cycle; secondary platforms are context only.

Core Question: "Did this content cycle, on its primary platform, create measurable signal strong enough to keep / discard / watch / block — and what should the next strategy/execution skill know?"

Why, methodology, history: references/playbook.md [PLAYBOOK]. Capability metadata (route triggers, prerequisites, load map): routing.yaml.

Critical Gates

1. Existing eval loop required. program.md + context.md absent → NEEDS_CONTEXT, recommend /run-pipeline. This skill does not create loops.
2. Organic non-video content only. Text / image / carousel posts. Short-form video → evaluate-shortform; paid-ad → evaluate-ad. Otherwise → NEEDS_CONTEXT, route to sibling.
3. Measurement evidence + primary metric decides the row. Need ≥1 metric source, window, and current value for the loop's primary metric (engagement rate · save rate · CTR · conversion rate — operator-pick via program.md). Secondary metrics explain diagnosis; they don't override unless program.md defines a guardrail failure.
4. One primary platform per cycle. Secondary platforms live in Cross-Platform Context subsection — they inform diagnosis but DO NOT drive the verdict. A 9-platform campaign = 9 separate cycles, one primary platform each.
5. No fabricated analytics. Unknown values stay unknown. Manual notes only when labeled operator-supplied + tied to date/window/source.
6. Attribution confidence must be explicit. Every verdict includes sample size (impressions/reach + window), baseline comparability (same platform, same content type, comparable window), confounders (algorithm change, posting-time shift, follower-count change, cross-post cannibalization), and confidence: high | medium | low | blocked.
7. Evaluation does not generate content. Recommend changes; route copy authorship to write-social, distribution to publish-social, visual assets to produce-asset.

Responsibility Split

/run-pipeline owns loop setup + program.md / context.md / results.tsv schema + durable learnings. This skill owns post-publish organic-content evidence snapshots scored against one primary platform. /write-social owns next-cycle copy. /evaluate-shortform + /evaluate-ad own their lanes.

Channel store — single write direction. This skill's metric-ingest agent is the sole writer of the cross-initiative channel store (fully-keyed Metric Packets → .forsvn/performance/<platform>.tsv + ledger advance; see the agent's Channel Store Append step). Diagnosis may read it for trend context — bun scripts/query-performance.ts <platform> --json (references/_shared/performance-data.md) — but never opens a second append path.

Inputs

Required: loop slug/path · primary-platform tag (linkedin · instagram · x · facebook · threads) · source write-social artifact (docs/forsvn/artifacts/marketing/copy/[platform]-[date]-[slug].md) · measurement window · primary metric value + source · reach/impressions (sample-size floor).

Recommended: baseline/prior-cycle row · engagement breakdown (likes/saves/shares/comments split — drives Engagement-Quality dim) · click-through + conversion · qualitative evidence (comment sentiment, replies, DMs — honest, not fabricated) · secondary-platform headline metrics (Cross-Platform Context) · guardrails from program.md.

Outputs

.forsvn/loops/[slug]/evals/YYYY-MM-DD-cycle-N.md + append one row to results.tsv via bun scripts/append-loop-result.ts (8-col helper) + update learnings.md ONLY for high-confidence keep/discard lessons (critic-gated) + run bun scripts/manifest-sync.ts.

Agent Manifest + Dispatch

4 sub-agents: Layer 1 parallel (Metric Ingest + Diagnosis) → Layer 2 (Recommendation) → Layer 3 (Critic). Critic FAIL → revise once; still FAIL → no ledger row + BLOCKED. Full agent table + per-layer dispatch + 7-dim rubric: references/agent-manifest.md. Domain rubric: references/rubric.md. Shared frame: _shared/evaluation-loop-rubric.md.

Pre-Dispatch

Canonical: _shared/pre-dispatch-protocol.md + _shared/eval-loop-spec.md. Hard-blocks (BEFORE Cold Start): missing program.md/context.md → NEEDS_CONTEXT + /run-pipeline; content is short-form video → route to evaluate-shortform; no measurement evidence OR missing primary-platform tag → BLOCKED; custom 10+ col results.tsv → warn + hand-edit. Cold Start: 6 bundled questions (loop · primary platform · source write-social artifact path · window · primary metric value/baseline · reach). Full read-order + templates + --fast behavior: references/procedures/pre-dispatch.md [PROCEDURE]. Session execution profile (single-vs-multi): inherit per references/_shared/execution-policy.md.

Artifact Contract

• Path: .forsvn/loops/[slug]/evals/YYYY-MM-DD-cycle-N.md. Lifecycle: evaluation.
• Frontmatter (10 fields): skill / version / date / status / summary / purpose / lifecycle / use_when / do_not_use_when / upstream / downstream + provenance (input_artifacts = source write-social + brand/BRAND.md + research/icp-research.md; output_eval: null).
• Body sections (8): Title · Verdict · Evidence (6-col table) · What Changed This Cycle · Diagnosis (Likely Drivers + Engagement-Quality Signals + Cross-Platform Context + Confounders) · Next Cycle Recommendation · Results Row (8-col TSV) · Learning Promotion. Verdict block must name the primary platform explicitly (Gate 4); Evidence table scopes metrics to that platform.
• Results Row (8 cols): cycle date artifact primary_metric value baseline status description. status ∈ keep|discard|watch|blocked (Critic Hard Fail otherwise); description includes the primary-platform tag.
• Cross-stack contract: consumed by future content-eval cycles (trend analysis) + write-social --rev=N+1 (hypothesis seeding) + human reviewers. Schema changes require atomic update across _shared/eval-loop-spec.md + downstream callers.

Full template + Evidence/Engagement-Quality/Results/Learning formats + helper invocation: references/format-conventions.md [PROCEDURE].

Results Row Helper

bun scripts/append-loop-result.ts "<loop slug>" \
  --artifact evals/YYYY-MM-DD-cycle-N.md \
  --metric "<primary metric>" --value "<current>" --baseline "<baseline>" \
  --status "<keep|discard|watch|blocked>" --description "<one sentence — include primary-platform>"

Do not append on Critic FAIL — return BLOCKED instead.

Critic Override Protocol

Operator ships despite critic FAIL (or accepts pass-with-concerns) — log BEFORE writing artifact or ledger row: bun scripts/log-critic-override.ts --skill evaluate-content …. Three overrides → rubric-revision escalation. An override never promotes a contested cycle to keep; a no-override FAIL still returns BLOCKED. Full protocol: references/_shared/critic-override-protocol.md [PROCEDURE].

Anti-Patterns

references/anti-patterns.md [ANTI-PATTERN] — content-eval rows + 4 cross-cutting marketing-stack rows. Most common: vanity-metric inflation (Gate 3 + Critic "Engagement-Quality Discrimination"), cross-platform contamination of the verdict (Gate 4 + Critic "Platform-Fit"), scope drift to write-social (Gate 7 + Critic "Decision Discipline"), missing source write-social artifact (Critic Hard Fail).

Worked Example

Organic content cycle (engagement-quality vs vanity, algorithm-spike → watch, critic PASS): references/examples/content-eval-cycle-walkthrough.md.

Durable Rules (protected)

<!-- SLOW_UPDATE_START --> <!-- No pinned rules yet. Populate via the slow-update workflow (see references/slow-update-fence.md). Each pinned rule must (a) be procedural not instance-specific, (b) be earned from a regression or critic-flagged failure, (c) cite the artifact / decision record that justified pinning. --> <!-- SLOW_UPDATE_END -->

Completion Status

• DONE — eval artifact written, ledger row appended, critic PASS.
• DONE_WITH_CONCERNS — artifact + row written, but confidence low/medium or confounders material.
• NEEDS_CONTEXT — missing loop, source write-social artifact, primary-platform tag, or required metric evidence; OR content is short-form video / paid ad (route to right sibling).
• BLOCKED — contradictory data, no measurement evidence, filesystem failure, or critic failed after revision.

References

• references/{playbook, agent-manifest, rubric, format-conventions, anti-patterns}.md + procedures/{pre-dispatch, dispatch-mechanics}.md
• _shared/{eval-loop-spec, evaluation-loop-rubric, pre-dispatch-protocol, critic-override-protocol, quality-dashboard-spec}.md
• Siblings: write-social (downstream), run-pipeline (loop scaffolding), evaluate-{ad, shortform, landing-page, campaign} (sibling lanes)