Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add genudo-ai/genudo_mcp --skill mine-customer-chatsgit clone --depth 1 https://github.com/genudo-ai/genudo_mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/genudo-ai/genudo_mcp/mine-customer-chats)<a href="https://agentmods.dev/skills/genudo-ai/genudo_mcp/mine-customer-chats"><img src="https://agentmods.dev/badge/skills/genudo-ai/genudo_mcp/mine-customer-chats/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/genudo-ai/genudo_mcp/mine-customer-chats"><img src="https://agentmods.dev/badge/skills/genudo-ai/genudo_mcp/mine-customer-chats.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00097 | $0.00899 |
| Opus 5.5 | $0.00039 | $0.00360 |
| Sonnet 5.5 | $0.00019 | $0.00180 |
| Haiku 4.5 | $0.00010 | $0.00090 |
Grade A, and why
mine-customer-chats scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Mine Customer Chats
Turn real conversations into an empirical picture of demand: what customers ask, how often, and where the agent's answers or knowledge run out. Read-only — nothing is written to the account.
1. Scope the sample
- Confirm the pipeline (
list_pipelines) and a window — 14–30 days is usually enough. - Pull customer-side messages with
list_messagesfiltered bypipeline_id,start_date,end_dateandrole. Fetch one unfiltered page first to learn the role values this account returns, then filter to the customer side. Page withnext_cursor. - Skip test traffic: opportunities with
is_test,@example.comemails, or names ending inTEST R<n>(list_opportunitiesshows these). - Stay under the 30 requests/minute limit; on error
-32029or HTTP 429, wait ~25s and resume.
2. Clean the turns
- Keep customer text only. Merge consecutive customer messages sent within ~60s into one turn.
- Drop pure greetings, stickers, and empty messages.
- Keep dialect and transliteration as written (Egyptian/Gulf Arabic, Franco-Arabic like
bkam,3ayez) — those spellings are exactly what tests must exercise later.
3. Cluster intents
Derive clusters from the data and the pipeline's business, not a fixed list. A typical commerce set, to adapt:
| Cluster | Examples |
|---|---|
| Price & offers | price, bundles, discounts, free-shipping threshold |
| Product fit | which variant suits me, sizes, ingredients/specs |
| Usage | how to use, compatibility, results |
| Order intent | "I want to order", sends address/phone |
| Payment | cash on delivery, wallets, cards, instalments |
| Delivery | time to my city, shipping cost, tracking |
| Post-purchase | delays, changes, cancellation, returns, complaints |
Count each cluster's share of turns and of conversations, and keep 3–5 verbatim examples.
4. Find the gaps
For a sample of conversations per cluster, read the agent's replies around the customer turn and flag:
- Deflected or vague answers ("I'll check and get back to you") to a factual question.
- Knowledge gaps — run
search_knowledge_tablewith the customer's question; no relevant row means a missing fact. Note the exact question and the answer the business should give. - Premature action — an order or webhook fired before the customer confirmed (cross-check
list_actionsfor the pipeline's actions and the opportunity's stage history). - Missed action — the customer clearly confirmed but nothing fired or the stage didn't move.
- Wrong stage — the conversation's intent changed but the opportunity stayed put
(
list_pipeline_stagesfor the expected route).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 77 lines · 97 tokens per session scan A 0175c2bebbb6
mine-customer-chats is a skill published in the GitHub repository genudo-ai/genudo_mcp (0 stars, last pushed 6d ago), licensed MIT. It adds 97 tokens to every session and 899 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-11.
Other skills, from other repositories
sdlc-accelerate
End-to-end SDLC ramp-up from idea to construction-ready with automated phase transitions.
ralph-memory
Manage Al semantic memory entries — list, query, and clear lessons learned across loop iterations.
config-validator
Validate AIWG configuration files and project setup for correctness and completeness.
project-awareness
Detect project type, AIWG framework state, team configuration, and active work to summarize status and recommend next actions.
summarize-transcript
Analyze and summarize a transcript, meeting notes, or discussion thread into a clear actionable document.
validate-addon
Validate an entire AIWG addon package for completeness and release readiness.