| name | daily-ai-intelligence |
| description | Run the daily AI technical-intelligence briefing workflow. Polls governed registry sources (GitHub releases, RSS/Atom, Hacker News, Reddit, Bluesky, X, papers, video/audio, manual items), normalizes and deduplicates events, ranks with topic memory and explicit feedback, generates a validated daily Markdown brief, and optionally exports to Obsidian, builds hot-topic dossiers, records feedback, runs weekly trend synthesis, or compares ranked output across registries. Use when the user asks for a "daily AI brief", "AI tooling watch", "coding-agent update scan", "MCP/Copilot/Claude Code daily brief", "weekly AI tooling trend memo", "hot-topic dossier from a daily brief", a "weekly AI tooling synthesis", or to "rank/validate/export today's AI intelligence". Do NOT use for academic literature reviews, PDF download/conversion, citation-graph expansion, or paper-only synthesis (use the `research-pipeline` skill instead).
|
| license | MIT |
Daily AI Intelligence
When To Trigger
- "Run today's AI brief" / "daily AI technical brief" / "AI tooling watch"
- "Coding-agent update scan" / "MCP daily brief" / "Copilot/Claude Code digest"
- "Weekly AI tooling trend memo" / "Weekly synthesis from my daily briefs"
- "Generate a hot-topic dossier for cluster X"
- "Record feedback on this cluster/source/topic"
- "Export today's brief to my Obsidian vault"
Do not trigger for academic literature reviews, paper-only synthesis,
PDF conversion, general web search, or social-media scraping outside the
reviewed registry. Do not use this skill for academic research — hand off
to research-pipeline (academic) as needed.
Launch
Always launch through the manifest-governed runner. Never bypass it.
SKILL_DIR=~/.claude/skills/daily-ai-intelligence
REG=~/my-daily-registry.toml
WS=./workspace/briefing
python3 $SKILL_DIR/runners/runner.py --registry "$REG" --workspace "$WS"
The runner reads manifest.json, initialises workflow_state.json under
<WS>/<date>/, and drives all tasks in dependency order via the
research-pipeline brief CLI/MCP tools. Optional tasks
(review-ranked, dossier, feedback, export-obsidian, weekly-synthesis) are
triggered only when the corresponding context key is set (via --reviewer,
--cluster-id, --vault, --week). Completion is proved by artifact
existence + schema validation — not by agent claim.
Rules
- Do not bypass the runner. Never call
brief CLI commands directly
without the orchestrator updating workflow_state.json.
- Resume = re-run the runner. Pass
--state <existing>.json to
continue an interrupted workflow. Idempotent: accepted tasks are skipped.
- Reviewer gates. If the optional rank_reviewer returns
status: "rejected",
fix ranked events, reset the task to pending, and re-run. Do not
override a rejected verdict.
- Validate before delivery. Never surface a brief whose
validation/validation.json reports failures. Low-signal days are
valid outputs — do not pad them.
- Registry-only sources. Never add ad-hoc scraping. New sources
require an explicit reviewed registry entry.
- Privacy gate. Never send raw, unclustered source dumps to a
cloud LLM. Only ranked evidence packs and validated reports may
leave the local workspace.
References
| File | Load when |
|---|
references/workflow-steps.md | Per-agent SKILL_DIR paths, or understanding/diagnosing an orchestrated task |
references/command-reference.md | All brief subcommands, options, MCP tool map |
references/source-policy.md | Adding/enabling a source or refusing expansion |
references/feedback-loop.md | Recording feedback or computing preference adjustments |
references/report-templates.md | Writing or validating a daily/weekly report |
references/agent-evaluation.md | Evaluating or demonstrating the agent |
references/troubleshooting.md | Empty reports, validation failures, dedup/vault errors |
Final Response To User
When workflow_state.json shows status: complete:
- Show the brief path and the validation result summary.
- List the top 5 clusters and their event counts.
- Offer to build a dossier, export to Obsidian, or run weekly synthesis.
