CO2··· · CH4··· · TEMP_ANOM··· · ARCTIC_ICE··· · FIRES_24H··· · QUAKES_M4··· · ENSO··· · CORAL_DHW··· · DEFOR_AMZ··· · CARBON_EU··· · CO2··· · CH4··· · TEMP_ANOM··· · FIRES_24H··· · ENSO···
CL> PIPELINE 10 STAGES · FETCH → REFLECT

The agent runs an autonomous content pipeline. Every published post survives ten stages — never a shortcut. Each stage has its own persona pool, its own quality gate, its own kill-switch. Defaults are safe; the operator can interject at any stage.

STAGES
10
INGEST → LEARN
ADAPTERS
23
ACADEMIC/RSS/SOCIAL/MARKETS
CONFIGS
96
FEEDS + QUERIES
COUNCIL
15
VOTE @ S07 REVIEW
QUERIES
141
DRIVE S01 FETCH
DAY_CAP
3
MAX POSTS / DAY

▸ Stage Flow

#STAGEVERBPURPOSETARGET
01 Fetch ingest Pull from 96 source configurations across 23 adapter types — academic papers (arXiv, PubMed, Crossref, bioRxiv), 64 RSS feeds, 41 X/Twitter queries, 30 subreddits, satellite data, real-time environmental telemetry, carbon markets, ESG filings, policy feeds. 5 min
02 Curate extract Extract verifiable claims from raw items, attach evidence ratings (strength, novelty, replication status), deduplicate against the knowledge graph, and tag with topic taxonomy. 3 min
03 Rank prioritise Score every claim against the scoring-weights config — evidence strength, novelty, audience fit, topic diversity, deadline urgency — then surface the top candidates for the draft stage. < 1 min
04 Draft compose Generate platform-specific draft posts (X, LinkedIn, Bluesky) with the science-writer persona. Every claim cites its source with a verifiable link. No banned phrases, no overclaim. 2 min
05 Verify fact-check Run the evidence protocol: every numeric claim must trace to a source; every phrase passes the compliance guard; every link resolves; nothing exceeds the platform character budget. 1 min
06 Revise refine Persona-driven revision loop. Editor, audience-reviewer, engagement-optimizer iterate on the draft. Reviser merges proposals into a single canonical version. 2 min
07 Review deliberate The Council convenes — 15 content personas vote, propose changes, and request more discussion. Operator can interject or set chair. No post advances without quorum + approval. 5 min
08 Publish ship Send to the platform publishers (X, LinkedIn, Bluesky) with idempotency guards. Safety-gated by default: approval mode ON, dry-run ON, ECA_DISABLE_PUBLISH kill-switch, max 3 posts/day. < 1 min
09 Interact engage Track replies, mentions, citations of the published post. Capture sentiment, correction requests, engagement metrics. Feed back into ranking and learning. continuous
10 Reflect learn Outreach + reflection. Reach out to experts whose work was cited, log feedback, update topic trust scores, adjust personas if council voted on shifts, write the transparency report entry. continuous

▸ SAFETY GATES

DEFAULT ON
ONApproval ON
ONDry-run ON
ONECA_DISABLE_PUBLISH
ONMax 3 posts/day
ONMon-Fri windows only
TO_PUBLISH: operator must explicitly disable the relevant guard and acknowledge the consequence. Captain refuses to ship by default.

▸ ARCHITECTURE

WHERE STAGES RUN
S01-S03SQLite + 23 source adapters · edge-cached
S04-S06Pluggable LLM client · Anthropic + OpenAI
S0715-persona council deliberation engine
S08Publisher registry · X / Bluesky / LinkedIn
S09-S10Continuous loops · learning back into Rank