Agent-Reach Ops Source Map

Status: public-safe field pattern for connector-backed content operations.

This note turns the Agent-Reach value-explorer experiments into one reusable LoopX source profile inside the broader connector source map. It is not a hard dependency on Agent-Reach and it does not grant permission to publish. It describes how a LoopX agent can use Agent-Reach as a source router, then let LoopX own evidence, gates, drafts, monitors, and handoff state.

Agents do not need to read this document before acting. The executable entrypoint is:

loopx value-connectors source-map --connector agent_reach_ops_source_map --format json

For all currently surfaced connector profiles, use:

loopx value-connectors source-map --format json

When To Use It

Use this pattern when a LoopX agent needs to:

• find public signals for a post, reply, launch note, or contributor-facing update;
• compare whether an idea is a mature category, an emerging phrase, or weak noise;
• build a source-backed draft without copying raw timelines, private chats, credentials, or local logs into LoopX state;
• keep publishing auditable even when the owner grants broad posting discretion.

Do not use it for credential collection, account setup, private timeline dumps, mass outreach, captcha bypass, trading, paid data, or production actions.

Operating Loop

Every connector-backed content run should use this sequence:

doctor -> route selection -> read-only source map -> maturity scoring
       -> ops brief -> draft packet -> publish/audit gate -> compact monitor

The important rule is that Agent-Reach is the source router, not the source of truth for action. LoopX keeps the compact action contract:

• what source was read;
• whether the boundary is public, logged-in read-only, private-needs-review, or forbidden;
• what claim the source supports;
• what draft was produced;
• what publish authority exists;
• what monitor or stop condition follows.

Route Selection

Start with:

agent-reach doctor --json

Treat the doctor output as capability evidence. A route is usable only when the active backend is available and the access boundary is clear.

Suggested route mapping:

Route	Typical backend	Boundary	Safe use
GitHub	gh CLI	logged-in read	repo/search metadata, public stars, descriptions, issues/PR metadata when explicitly routed
Web	Jina Reader	public no-login	public docs and articles, with light quoting only
RSS	feedparser	public no-login	feed titles, links, timestamps, summaries
V2EX	public API	public no-login	hot topics and public replies as community signal, usually monitor-grade
Bilibili	public search API or bili-cli	public no-login	video title, author, play count, public URL
X/Reddit/Xiaohongshu/Facebook/Instagram	platform CLI or OpenCLI	logged-in read	read-only public or account-visible metadata only; no posting without a separate publish action record

If a route needs browser cookies, platform login, private groups, DMs, account setup, or raw body expansion, stop at metadata-only and project a gate.

Evidence Card Shape

Agents should emit compact cards before drafting:

agent_reach_ops_signal_v0:
  channel: github | web | rss | v2ex | bilibili | x | reddit | xiaohongshu | other
  backend: gh CLI | Jina Reader | feedparser | V2EX API | Bilibili public search API | ...
  query: string
  title: string
  url: string | null
  summary: string
  boundary: public_no_login | logged_in_read | private_needs_review | forbidden
  operation: read
  observed_at: ISO-8601 timestamp
  confidence: doctor_status | source_metadata | source_body_reviewed
  maturity_score: 0 | 1 | 2 | 3
  maturity_reason: string

operation must be read. A connector run that creates external_write cards is invalid for this source-map stage.

Maturity Scoring

Keep the scoring deliberately simple so the next agent can reuse it:

Score	Meaning	Example signal
0	Noise or unavailable route	unrelated hot topic, route missing, or stale source
1	Weak exploratory signal	exact phrase appears but with little adoption
2	Emerging signal	repeated usage, modest stars, replies, or public attention
3	Mature signal	strong adoption, many stars/views/replies, or multiple independent sources

For public GitHub search, stars can be a first-pass proxy:

• >= 1000: mature category signal;
• >= 100: emerging category signal;
• >= 10: weak but visible;
• < 10: exploratory unless other sources corroborate it.

For public video/community sources, use attention only as a clue. Do not copy unverified claims from drama, rumor, or account-risk videos into LoopX claims.

Ops Brief

Before drafting, produce a short brief:

ops_brief:
  source_batch: evidence-card file or packet id
  mature_signals:
    - claim:
      supporting_cards:
      why_it_matters_for_loopx:
  weak_signals:
    - claim:
      reason_to_monitor_instead_of_draft:
  recommended_angles:
    - audience:
      body_angle:
      evidence_refs:
  stop_conditions:
    - account boundary unclear
    - source only supports metadata, not body quote
    - post would repeat unverified platform drama

The brief is the handoff point. A different agent should be able to draft or review from it without reading raw connector output.

Content-Ops Drafting

A draft is valid only when it has:

• a named angle and target reader;
• a source map with public/private status;
• exact body text;
• media plan;
• repo or docs link when the post is about LoopX;
• account/channel/timing record when publishing is allowed;
• stop condition and first monitor plan.

Broad owner permission may allow an agent to publish according to judgment, but it does not remove the audit requirement. The publish record still needs the final body, active account or channel, source refs, timestamp, and follow-up monitor boundary.

Reusable Agent Prompt

When onboarding a new LoopX agent for creator/operator work, the preferred instruction is to call the CLI packet:

Run `loopx value-connectors source-map --format json` before drafting from
external signals. Choose a read-only source profile, emit compact evidence
cards, score maturity, and write an ops brief. Use `loopx value-connectors plan`
before any signup, send, post, reply, upload, production action, credentialed
read, or private-source expansion.

The longer fallback prompt is:

Before drafting social content, run connector-first source mapping.
Use Agent-Reach routes only for read-only signal collection unless a separate
LoopX gate authorizes a publish action. Emit compact evidence cards, score
maturity, write an ops brief, and draft from the brief. If publishing is
authorized, record the exact body/account/time/source-map/stop-condition before
posting. If the active account or source boundary is unclear, stop with a
no-send packet.

Example Finding

A value-explorer run using Agent-Reach routes found:

• mature GitHub signals for long-running AI agent, AI agent control plane, and agent loop engineering;
• high Bilibili attention around Claude Code and AI Agent setup/tutorial content;
• weak V2EX hot-topic relevance for this exact LoopX angle at that moment.

The reusable conclusion is:

Connector-backed agents can find the trend. LoopX should make the action
reviewable: evidence cards, gates, draft packets, and monitors.

Productization Boundary

The first stable part is now productized as a packet:

• loopx value-connectors source-map --connector agent_reach_ops_source_map ...;

Keep live collection and publish tooling local until at least two successful batches prove the card shape and source boundaries. Productize only the stable parts:

• loopx content-ops draft --from-source-map ...;
• loopx content-ops publish-record --from-draft ...;
• loopx content-ops monitor --published-url ....

Do not productize raw provider payload retention, platform-specific cookies, or publish shortcuts.