| name | x-intelligence-reporting |
| description | Summarize X/Twitter intelligence digests for ML, AI, frontier/foundation model practice, agents, evals, tool use, harness engineering, loop engineering, and loopcraft. Use when asked to run or interpret the loopcraft X intelligence workflow, prepare a daily Codex automation report from X posts, highlight frontier-lab posts, discover follow candidates, or extract direct links from X posts and linked external articles/blogs/papers. |
X Intelligence Reporting
Workflow
- Work in the loopcraft source tree from the runtime context (the repo with
Makefile, config/, and src/) unless the user gives another workspace.
- Fetch a fresh digest by running the direct CLI:
PYTHONPATH=src python -m loopcraft.research_intel.x.cli run --config config/x_intel.yaml.
You are executing inside a headless loop run, so do not re-enter the control
plane for this loop (no loopctl run / make run of x-intel) — that
recurses. (The control plane also enforces this programmatically: it sets
LOOPCRAFT_ACTIVE_LOOP for the run and refuses re-entry into the same
loop while it is set.) If network access or X API credentials fail, report that first.
- Read
state/research/x/latest.md and latest.json from the loopcraft memory ledger (or the emitted direct-CLI paths). Prefer JSON for exact fields and Markdown for human-readable ordering.
- When useful, run
PYTHONPATH=src python -m loopcraft.research_intel.x.cli discover-follows --config config/x_intel.yaml and read the emitted follow-candidate Markdown/JSON paths.
- When summarizing, include both the X post permalink and any external links from
entities.urls[*].expanded_url or Markdown External links: lines.
- Group findings by theme rather than only by score. Use these default themes:
- Frontier-lab signal
- Loop engineering and long-running agents
- Harnesses, orchestration, verifiers, and memory
- Evals, observability, traces, and post-training
- Tool use, local systems, and production practice
- Highlight frontier-lab posts separately. Treat frontier-lab attribution as explicit handle/config metadata, not loose profile-text inference.
- Include follow recommendations when available. For each candidate, include why they surfaced and at least one evidence post link.
- Call out practical loopcraft implications: what to read, what to implement, what to watch, and what should influence automation or agent-harness design.
Reporting Standards
- Lead with errors, rate limits, auth failures, or empty fetches if present.
- Mention generated date and fetched/ranked counts when available.
- Link every cited X post using
https://x.com/{username}/status/{id}.
- Link external resources explicitly by title or domain; do not hide important blog/paper links behind the X post alone.
- Avoid over-weighting repeated posts from one author; collapse near-duplicates in prose.
- Be clear when a conclusion is inferred from post text rather than verified from the linked article.
- Keep the report concise enough for daily reading, but include enough links that the user can immediately open the primary sources.
Useful Commands
In-loop, always use the direct CLI below; never re-run this loop through the
control plane (that recurses). Operators trigger the loop from a shell via its
documented make/loopctl target — this skill should not.
PYTHONPATH=src python -m loopcraft.research_intel.x.cli run --config config/x_intel.yaml
PYTHONPATH=src python -m loopcraft.research_intel.x.cli snapshot-following
PYTHONPATH=src python -m loopcraft.research_intel.x.cli discover-follows --config config/x_intel.yaml
PYTHONPATH=src python -m pytest -q
Expected Output Shape
Use this structure unless the user asks for something else:
**Daily X Intelligence**
Fetched N posts; ranked M. Errors: none.
**Main Findings**
- Theme: concise synthesis. X post: [author](...). External: [resource](...). Why it matters: ...
**Frontier-Lab Highlights**
- Lab/person: summary. X post: [...](...). External: [...](...).
**Follow Candidates**
- Candidate: why they surfaced. Profile: [...](...). Evidence: [...](...).
**Loopcraft Implications**
- Actionable observation for loop engineering, harnesses, evals, agents, or tool use.