AI Self-Improvement Digest Generate daily digests that help AI agents improve their own reasoning, execution, and capabilities. The skill reviews it's research against it's own setup and makes recommendations for improvement every day. This skill is leverages both the Brave Search and X API (special shoutout to @frankdegods x-research skill) though both are not required for it to run. Description This skill creates a daily digest focused on AI self-improvement material, not general AI news. The goal is to surface content that helps an AI agent get better at its job by learning from:

Harness and system prompt engineering patterns Skill and tool development approaches Self-evaluation and debugging techniques Multi-agent coordination strategies Memory and context management Task decomposition and workflow automation Reasoning and execution patterns Use When Setting up daily learning routines for AI agents Building an AI agent that improves itself over time Curating educational content for agent development Creating a structured self-improvement loop Don't Use When Looking for general AI/tech news Need breaking announcements or model releases Want business or market news Setup

1. Configure Brave Search API Add to your Clawdbot config:

{ "tools": { "web": { "search": { "enabled": true, "provider": "brave", "apiKey": "YOUR_BRAVE_API_KEY" } } } }

Get a Brave Search API key at: https://brave.com/search/api/ 2. Create Tracking File Create memory/ai-digest-posted.json:

{ "posted": [], "experiments": [], "skillsEvaluated": [] } 3. Schedule the Cron Job

Via Clawdbot CLI or cron tool

clawdbot cron add --name ai-self-improvement-digest
--schedule "30 8 * * *"
--tz "America/New_York"
--message "[paste the prompt below]" Cron Job Prompt Generate a daily AI Self-Improvement Digest and deliver to your configured channel.

PURPOSE: Surface content that helps you (the AI) improve your own reasoning, execution, and capabilities. This is not news - it's training material for self-improvement.

STEP 1: DEDUPLICATION (MANDATORY) Read memory/ai-digest-posted.json. Skip anything already posted (by URL or substantially similar topic).

STEP 2: SCAN SOURCES Use web_search and web_fetch to check these sources for content from last 24-72h:

Tier 1 (daily):

• Anthropic Engineering: anthropic.com/engineering (harnesses, evals, multi-agent)
• Simon Willison: simonwillison.net (practical patterns, tool commentary)
• Geoff Huntley: ghuntley.com (agent philosophy, MCP patterns)
• X/Twitter: Real-time practitioner insights via x-research skill (see below)
• Hacker News: news.ycombinator.com AI/agent threads (high signal discussions)
• Lilian Weng: lilianweng.github.io (deep technical AI posts, agent architectures)

Tier 2 (2-3x/week):

• Latent Space: latent.space (industry depth, interviews)
• Cursor Blog: cursor.com/blog (coding agent patterns)
• David Crawshaw: crawshaw.io (agent experience reports)
• Mitchell Hashimoto: mitchellh.com (practical engineering)
• Eugene Yan: eugeneyan.com (ML systems, production patterns)
• Chip Huyen: huyenchip.com (ML systems design, practical deployment)

Tier 3 (weekly scan):

• arXiv cs.CL/cs.AI: search for 'agent reasoning tool use'
• GitHub trending: AI agent repos, MCP servers
• Hacker News: AI coding/agent threads

X/Twitter Research (Tier 2)

When scanning X, use the x-research skill:

bash cd ~/clawd/skills/x-research bun run x-search.ts search "AI agents harness" --quick --quality bun run x-search.ts search "MCP server" --quick --quality bun run x-search.ts search "from:simonw" --quick bun run x-search.ts search "from:geoffreylitt" --quick

Focus on:

Practical implementation insights (not hype) Harness/prompt engineering patterns Tool and skill development discussions Multi-agent coordination experiences Links to detailed writeups (then fetch with web_fetch)

STEP 3: FILTER FOR SELF-IMPROVEMENT RELEVANCE Only include items that help improve capabilities in:

Harness/system prompt design Skill and tool development Self-evaluation and debugging Multi-agent coordination Memory and context management Task decomposition and workflow automation Reasoning patterns

EXCLUDE: General AI news, model announcements, business news, ethics debates, items already in ai-digest-posted.json.

STEP 4: FORMAT (3-5 items) For each item:

[Title] — [Source] What: [1-sentence summary] Why it matters for self-improvement: [How this helps you get better] Takeaway: [Specific pattern, technique, or experiment to try] Relevance: [⭐ to ⭐⭐⭐⭐⭐]

STEP 5: EXPERIMENT SUGGESTION 💡 Today's experiment: [One small thing to try based on the digest that could improve your capabilities]

STEP 6: SETUP REVIEW (MANDATORY) Review the content you just surfaced against your existing setup (AGENTS.md, TOOLS.md, skills/, cron jobs, memory patterns). Make concrete, affirmative suggestions:

🔧 Setup Review Based on today's findings:

Let's add [specific thing] because [reason tied to content found] Let's update [existing thing] to [improvement] because [reason]

If nothing is actionable: "No changes needed today — our current setup handles these patterns well."

Key principles:

Ground suggestions in what you already have Use affirmative voice ("let's do X") not passive ("could consider X") Connect each suggestion to a specific article/finding from the digest It's okay to have no suggestions if nothing is actionable

STEP 7: UPDATE TRACKING Append new items to memory/ai-digest-posted.json with date, title, url, topic.

FORMAT: 🧠 AI Self-Improvement Digest — [Date]

[Items...]

💡 Today's experiment: [...]

🔧 Setup Review [...]

📊 Feedback: 👍 = useful | 👎 = skip these | 🔥 = more like this | 💬 = thoughts

Deliver to your configured channel (Slack, Telegram, Discord, etc.).

Source Priority

Source	Priority	Focus
Anthropic Engineering	⭐⭐⭐	Harness design, evals, multi-agent
Simon Willison	⭐⭐⭐	Practical patterns, tools
Geoff Huntley	⭐⭐⭐	Agent philosophy, MCP
X/Twitter	⭐⭐⭐	Real-time practitioner insights
Hacker News	⭐⭐⭐	High-signal AI/agent discussions
Lilian Weng	⭐⭐⭐	Deep technical AI, agent architectures
Latent Space	⭐⭐	Industry depth
Cursor Blog	⭐⭐	Coding agent patterns
Eugene Yan	⭐⭐	ML systems, production patterns
Chip Huyen	⭐⭐	ML systems design
arXiv cs.CL/cs.AI	⭐⭐	Research foundations
GitHub Trending	⭐⭐	New tools, repos

Content Categories

1. Harness & System Prompt Engineering - How to structure agent instructions
2. Skill & Tool Development - New tools, MCP servers, integration patterns
3. Self-Evaluation & Improvement - How agents assess and improve themselves
4. Multi-Agent Coordination - Spawning, supervising, merging work
5. Memory & Context Management - RAG, long-term memory, compaction
6. Workflow Automation - Task decomposition, failure handling
7. Foundational Research - Academic work on agent capabilities

Entry Format

Each digest entry should include:

1. What it is - 1-sentence summary
2. Why it matters for self-improvement - How this helps the agent get better
3. Actionable takeaway - Specific pattern or experiment to try
4. Relevance score - 1-5 stars based on direct applicability

Setup Review Section

The Setup Review is a mandatory closing section that connects the day's findings to your existing infrastructure. It answers: "Based on what I learned today, should I change anything about how I operate?"

Good Setup Review Examples

✅ Affirmative, specific, grounded:

🔧 Setup Review Based on today's findings:

• Let's add a memory/experiments.md file to track harness experiments, since the Anthropic article showed experiment logging improves iteration speed
• Let's update the channel-monitor cron to include a self-check step before responding, based on the "pause and verify" pattern from Simon Willison's post

No changes needed for multi-agent coordination — our current sub-agent spawning pattern already follows the isolation principle discussed.

❌ Too vague or passive:

"Could consider maybe looking into some of the patterns mentioned" "Might be worth exploring memory improvements at some point"

When Nothing Is Actionable

It's perfectly fine to have no suggestions:

🔧 Setup Review No changes needed today — our current setup handles these patterns well. The memory compaction article validated our existing approach in AGENTS.md.

Self-Improvement Loop

The digest enables a continuous improvement cycle:

DAILY:

Read digest Pick 1 experiment to try Log outcome in memory/ai-digest-posted.json Review Setup Review suggestions with human

WEEKLY:

Review experiments Update harness/skills based on learnings Adjust source priorities based on value

Experiment Tracking

Extend memory/ai-digest-posted.json to track experiments:

json { "posted": [...], "experiments": [ { "date": "2026-02-16", "fromArticle": "effective-harnesses", "experiment": "Add checkpoint before sub-agent spawn", "outcome": "Reduced context loss by 40%", "learned": "Always checkpoint before spawning" } ], "skillsEvaluated": [ { "date": "2026-02-16", "skill": "mcp-postgres", "verdict": "useful", "notes": "Integrated for database queries" } ], "setupChanges": [ { "date": "2026-02-16", "change": "Added memory/experiments.md", "reason": "Track harness experiments per Anthropic article", "status": "implemented" } ] } Optional Enhancements X/Twitter Research (requires API key) Add the x-research skill for Twitter/X searches:

cd skills git clone https://github.com/rohunvora/x-research-skill.git x-research export X_BEARER_TOKEN="your-token"

Then add to Tier 2 sources:

X/Twitter: Search for AI agent discussions from key accounts Weekly Deep Dive Add a Friday cron for research paper reviews:

0 10 * * 5 # Fridays at 10am

Focus on 1-2 papers from arXiv with detailed summaries and implementation ideas. License MIT