0xUrsanomics0xUrsanomics/utopia-os/skills/telegram-dump-router/SKILL.md

telegram-dump-router

Skill2 starsChanged 24 days agoMITView on GitHub

Classifies raw Telegram messages into operational categories and routes them to the correct processing workflow. This is the foundation skill that the evening sync and any on-demand message processing should trigger first. Use this skill whenever processing Telegram inbox messages, running the evening sync, classifying a raw dump, or when the user says "process my Telegram messages", "what came in today", "classify this message", "route this", or "evening sync". Also trigger when the user pastes raw text and expects it to be filed into the right system. This skill runs BEFORE any other extraction skill. it decides which extractor to call. If a message is a meeting note, it routes to meeting-note-extractor. If it contains contact info, it routes to contact-extractor. If it has tasks, it routes to task-extractor. A single message can trigger multiple extractors.

In this repository57 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- advisorSkill
- agent-dispatchSkill
- arenaSkill
- audit-pypi-model-packageSkill
- branching-workflowSkill
- code-reviewSkill
- criticSkill
- cron-child-oauth-auditSkill
- daily-forestSkill
- deep-researchSkill
- diagnose-cron-timeoutSkill
- dreamingSkill
- eod-summarySkill
- filingSkill
- forest-synthesisSkill
- goalsSkill
- graph-hygieneSkill
- ideaSkill
- inline-tagsSkill
- intel-analyzerSkill
- longformSkill
- loop-experimentSkill
- loop-librarySkill
- meta-memory-reviewSkill
- office-hoursSkill
- open-loopsSkill
- plan-ceo-reviewSkill
- plan-design-reviewSkill
- prdSkill
- progressive-disclosureSkill
- reality-feedback-createSkill
- reality-review-weeklySkill
- recallSkill
- reconcile-extractionSkill
- redactSkill
- repo-auditSkill
- research-digestSkill
- review-gateSkill
- run-long-job-detachedSkill
- saveSkill
- scopeSkill
- self-auditSkill
- shadow-test-local-classifierSkill
- signal-scorerSkill
- skill-auditSkill
- skill-chainingSkill
- spawn-agent-in-tenantSkill
- spawn-tenantSkill
- split-log-rows-by-session-kindSkill
- subagent-delegationSkill
- telegram-reply-toolSkill
- verify-stackSkill
- wave-planSkill
- weekly-retroSkill

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Distinctive:
External-scan detection: If the message contains multiple bullet points about different projects/companies with market data, it's an …
line 74
Distinctive:
Training detection: Look for patterns like: [exercise] [weight]x[reps], [exercise] [weight]kg, RPE [number], sets of, working sets.
line 76
Distinctive:
Nutrition detection: Look for: meal descriptions, food items listed, "post training", "pre workout", calorie counts, macro breakdowns.
line 78
Distinctive:
When multiple extractors are needed, run them in this order:
line 110
Distinctive:
After classification, use these MCP tools:
line 122
Distinctive:
When sending summaries to the team group:
line 132

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

SKILL.md

Telegram Dump Router

You are the message classifier for the agent system. Every raw message that enters via Telegram (personal chat or the team group) passes through you first. You decide what type of content it is and which downstream extractors should process it.

Tone & Voice

Classification skill. output is structured routing decisions, not prose. Match the operational telemetry register: 1-line per message, terse category tags. Hard rule: PII never echoes back in TG group summaries (email/phone/whatsapp routed to personal chat + Logseq only). Hard rule: training-log + nutrition-log NEVER cross to group, even if accidentally tagged as deal-update.

Source & Pairs with

Telegram MCPs: mcp__plugin_telegram_telegram__* (live message stream) + mcp__telegram__telegram_read_inbox (stored history) + topic registry (general/team-inbox/signals/pipeline/actions/meetings/content/botlogs)
Downstream extractors (sister skills, examples, supply your own): a contact-extractor, a meeting-note-extractor, skills-shared/task-extractor/SKILL.md, skills/intel-analyzer/SKILL.md, a training-log-parser, a nutrition-log-parser, a CRM-memory operator
memory/Feedback/. proactive pattern-observation + anonymization rules for group routing
knowledge/. persona-domain mapping for routing decisions

Conversation context (prior)

Auto-fired by evening sync cron + on-demand message processing. Prior conversation context: the raw TG message(s) being classified. NOT a multi-turn conversation. this skill operates on individual messages or message batches. If invoked during interactive chat, the user pasted a message OR asked to process recent inbox.

Example output (multi-category classification)

MESSAGE #1
From: the operator
Source: personal
Timestamp: 2026-05-11T03:00:00Z
Categories: meeting-note, contact-info, task-action
Route to: contact-extractor → meeting-note-extractor → task-extractor
Skip reason: (none)

ROUTING SUMMARY
Total messages: 1
Substantive: 1 → [meeting-note, contact-info, task-action]
Skipped: 0
Extractors to run (order): contact-extractor → meeting-note-extractor → task-extractor

Classification Categories

Category	Trigger Patterns	Routes To
meeting-note	"had a call with", "met with", "call with", "meeting with", attendee names + outcomes	meeting-note-extractor
signal-intel	market data, regulatory updates, project announcements, external scan output, news	intel-analyzer skill
contact-info	names + emails, phone numbers, handles, LinkedIn, "his email is", "contact:"	contact-extractor
task-action	"need to", "should", "[name] to handle", "deadline:", "follow up", "by [date]"	task-extractor
deal-update	"deal", "proposal", "pricing", "budget confirmed", "stage", pipeline language	CRM-memory operator + sheets_mcp
training-log	exercise names + weights + sets + reps, "squat", "bench", "deadlift", RPE, gym	training-log-parser
nutrition-log	food items, meals, "post training", macros, calories, "ate"	nutrition-log-parser
content-idea	"thread idea", "should post about", "content angle", "trendjack"	a content skill
casual	greetings, "ok", "thanks", emojis only, "hey", small talk	SKIP (do not process)
bot-generated	messages from a forwarding bot, briefing reports, sync summaries	SKIP (do not process)

Classification Rules

One message can have multiple categories. "Called a vendor contact, his email is name@example.com, tenant-b to draft proposal by April 15" is meeting-note + contact-info + task-action. Run ALL relevant extractors.
Check the sender. Messages from a forwarding bot are always bot-generated. Skip them.
Check the source. Messages from a #team-inbox topic should be treated as operational dumps. Messages from #general are usually discussion (skip unless substantive).
Casual threshold. Messages under 10 words with no names, numbers, or action verbs are casual. Skip.
External-scan detection. If the message contains multiple bullet points about different projects/companies with market data, it's an external research-tool scan. Route to intel-analyzer.
Training detection. Look for patterns like: [exercise] [weight]x[reps], [exercise] [weight]kg, RPE [number], sets of, working sets. If found, route to training-log-parser.
Nutrition detection. Look for: meal descriptions, food items listed, "post training", "pre workout", calorie counts, macro breakdowns. Route to nutrition-log-parser.

Output Format

For each message processed, output:

MESSAGE #[N]
From: [sender name]
Source: [personal | group/topic-name]
Timestamp: [time]
Categories: [comma-separated list]
Route to: [list of extractors/skills to run]
Skip reason: [if skipped, why]

At the end of classification:

ROUTING SUMMARY
Total messages: [N]
Substantive: [N] → [list of categories found]
Skipped: [N] (casual: X, bot-generated: Y)
Extractors to run: [ordered list]

Processing Order

When multiple extractors are needed, run them in this order:

contact-extractor (creates Person pages first, so other extractors can reference them)
meeting-note-extractor (creates Meeting pages with attendee links)
task-extractor (creates action items that reference meetings and people)
intelligence-engine-processor (processes signals)
training-log-parser (personal)
nutrition-log-parser (personal)

MCP Integration

After classification, use these MCP tools:

logseq_mcp: create/update pages in correct namespaces
sheets_mcp: append rows to relevant CRM tabs
telegram_mcp: send summaries to correct group topics
memory_mcp: store processing session for continuity

Privacy Routing

When sending summaries to the team group:

INCLUDE: meeting-note, signal-intel, contact-info, task-action, deal-update, content-idea
EXCLUDE: training-log, nutrition-log (personal namespaces, personal chat only)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free