Skip to content

cgallic/kai-cmo-harness / legacy/content-report

content-report

Verified

Pull performance data for published content across every surface — LinkedIn, Dev.to, Google Search Console, GA4, and AI citation tracking (Perplexity, Anthropic, OpenAI). Grades each piece as winner, average, or underperformer and surfaces whether an AI search engine has cited it.

/content-report — The Analytics Lead

Pull performance data for your published content. Cross-references Google Search Console (position, CTR), GA4 (session duration), platform-native stats (LinkedIn impressions, Dev.to views), and AI citation checks (Perplexity, Anthropic, OpenAI web search) to grade each piece.

Preamble

source "$(dirname "$0")/../lib/preamble.sh"

Arguments

Optional:

  • --site: Filter by site key (e.g., kaicalls, connorgallic, meetkai)
  • --all: Show all published content (default if no args)

Example: /content-report --site kaicalls Example: /content-report

The Skill

Step 0: Fold In Any Legacy Log

The CANONICAL publish log is data/content_log.json (the learning loop — performance_check, pattern_extract, weekly_report — reads only this file). If a legacy ~/.kai-marketing/content-log.jsonl still exists, merge it first so its winners reach the learning loop (idempotent; safe to re-run):

python3 -m scripts.content.migrate_legacy_log
python3 -m scripts.self_improvement.performance_check --reconcile-only   # rebuild 30-day checks in data/pending_checks/

Step 1: Refresh Data

Before the report, refresh the data sources you care about. Skip any step where you don't have the API key.

# Dev.to — requires DEVTO_API_KEY (https://dev.to/settings/extensions)
python3 -m scripts.content.tracker_cli pull-devto

# AI citations — requires at least one of PERPLEXITY_API_KEY, ANTHROPIC_API_KEY, OPENAI_API_KEY
python3 -m scripts.content.tracker_cli citations

# LinkedIn is manual — record per article (interactive):
python3 -m scripts.content.tracker_cli linkedin --id <entry_id> --interactive
# Or non-interactive:
python3 -m scripts.content.tracker_cli linkedin --id <entry_id> \
    --url https://linkedin.com/pulse/... \
    --impressions 1850 --reactions 34 --comments 7 --reshares 2 --dms 3

Step 2: Load Enriched Report

python3 -m scripts.content.tracker_cli report --format json

If no published content is found, tell the user: "No published content in the log. Content is logged automatically when it passes /content-gate."

Step 3: Display Performance Overview

Present the data:

CONTENT PERFORMANCE REPORT
══════════════════════════════════════════

Total pieces: {N}
  Winners:         {green} {count} ({pct}%)
  Average:         {yellow} {count} ({pct}%)
  Underperformers: {red} {count} ({pct}%)
  Pending (< 30d): {gray} {count}

GRADING THRESHOLDS (search surface):
  Position ≤ 5  |  CTR ≥ 5%  |  Session ≥ 90s

RECENT CONTENT:
  [+] kaicalls-20260316   "AI receptionists"       pos:3  CTR:7.2%  91s   WINNER   li=4.2K/86r/3dm  cits=2/4
  [+] connor-20260414     "MeetKai audit"          —      —         —     PENDING  li=1.9K/52r/1dm  cits=1/3
  [~] kaicalls-20260310   "law firm answering"     pos:12 CTR:2.1%  45s   AVERAGE  devto=320v/12r   cits=0/3
  [-] abp-20260308        "backyard design"        pos:28 CTR:0.4%  22s   UNDER    li=80/1r         cits=0/3
  [?] kaicalls-20260322   "virtual receptionist"   (pending — 8 days old)

Legend for the compact metric strings:

  • li=1850/34r/2dm — LinkedIn impressions / reactions / DMs
  • devto=450v/12r — Dev.to views / reactions
  • cits=2/4 — positive AI citations / total citation checks run

Step 4: Insights

If there are 3+ graded pieces, provide insights:

  • Which format performs best (blog vs linkedin vs devto vs email)
  • Which persona performs best
  • Which hook type performs best (if tracked)
  • Average quality gate score of winners vs underperformers
  • AI citation patterns — which providers cite our content most often, which queries trigger citations, whether articles with high GSC position also attract AI citations

Example:

INSIGHTS:
  Best performing format:  linkedin (3/4 winners)
  Best performing persona: Shock Absorber (2/2 winners)
  Avg gate score winners:  82/100  vs  underperformers: 61/100

  AI CITATIONS:
    Most citing provider:  Perplexity (4 citations across 3 articles)
    Top-cited article:     connor-20260414 — 3 citations
    Queries that cite us:  "best AI answering service", "Connor Gallic AI",
                           "answering service alternatives 2026"
    Articles never cited:  abp-20260308, kaicalls-20260310  →  strong GEO-rewrite candidates

Step 5: Recommend Next Actions

Based on the data:

  • Pending 30-day review: "3 pieces ready for 30-day review. Performance data will be pulled on next check."
  • Underperformers: "Consider revising underperformers or running /content-retro to extract patterns."
  • Missing LinkedIn stats: "N LinkedIn pieces published in the last 14 days have no recorded metrics. Run kai-tracker linkedin --id <id> --interactive to record them."
  • Missing citation checks: "N tracked articles have no citations-targets.json entries. Use kai-tracker targets --id <id> --add-query '...' to configure."
  • No recent content: "No content published in the last 14 days. Run /content-ideas for topic suggestions."

Data Sources Behind This Skill

SourceWhere it livesRefreshed by
Publish log (canonical)data/content_log.json/content-write Step 6 / scripts.content.content_log.log_entry + mark_published
30-day pending checksdata/pending_checks/{id}.jsonAuto-created by content_log when a real URL is set; rebuilt by performance_check --reconcile-only
Legacy publish log~/.kai-marketing/content-log.jsonlRead-only — fold into the canonical log via scripts.content.migrate_legacy_log (Step 0)
Platform metrics (Dev.to, LinkedIn)~/.kai-marketing/metrics/{id}.jsonkai-tracker pull-devto / kai-tracker linkedin
AI citation events~/.kai-marketing/citations/{id}.jsonlkai-tracker citations (cron-compatible)
Citation targets~/.kai-marketing/citations-targets.jsonkai-tracker targets --id <id> --add-query '...'
GSC + GA4 gradingLive API calls via scripts/self_improvement/performance_check.pyInvoked automatically by kai-tracker report

Environment Variables

VariableRequired forHow to get it
DEVTO_API_KEYDev.to stat pullshttps://dev.to/settings/extensions
PERPLEXITY_API_KEYPerplexity citation checkshttps://www.perplexity.ai/settings/api
ANTHROPIC_API_KEYClaude web-search citation checksConsole API key (already required by the harness)
OPENAI_API_KEYOpenAI web-search citation checksExisting harness API key

If a key is missing, the corresponding check is skipped with a warning — the report still runs with whatever data is available.

Error Handling

  • No content log: Direct to /content-write to publish first piece
  • GSC/GA4 API unavailable: Show content log with platform/citation data, note "Search surface performance unavailable — showing platform metrics only"
  • All AI citation providers missing keys: Skip Step 1's citations call, note in the report "Citation tracking unavailable — set at least one of PERPLEXITY_API_KEY / ANTHROPIC_API_KEY / OPENAI_API_KEY"

Chain State

Reads from: data/content_log.json (canonical publish log), data/pending_checks/, ~/.kai-marketing/metrics/, ~/.kai-marketing/citations/ (platform/citation scratch) Feeds into: /content-retro (weekly pattern analysis, now including citation patterns)

Scheduling

Recommended cadence for a solo operator:

  • On publish: /content-gate writes to the log automatically
  • Weekly: kai-tracker pull-devto + kai-tracker citations (put in cron or weekly heartbeat)
  • Per LinkedIn post, 24h + 7d after publish: kai-tracker linkedin --id <id> --interactive
  • 30 days after publish: the existing performance_check cron handles GSC + GA4 automatically
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 193kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 193kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 193kUpdated Claude CodeCodexVerifiedInstall