vendor-insights

Generates faceted usage reports and HTML narratives for non-Claude agent session stores.

Updated May 31, 2026
One-click install
npx skills add https://github.com/4444J99/limen --skill vendor-insights-4444j99
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: vendor-insights
Source: https://github.com/4444J99/limen/tree/main/.claude/skills/vendor-insights
Command: npx skills add https://github.com/4444J99/limen --skill vendor-insights-4444j99

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Claude Code's built-in /insights only covers Claude sessions, leaving other agent lanes on the host (codex, copilot, opencode, antigravity/agy, gemini) without any equivalent usage reporting. This Skill produces the same class of per-session facets, aggregate narrative, and shareable HTML report for those vendors. ## Core Features & Use Cases - Cross-vendor session indexing: Builds a bounded session index per vendor with honest population metadata (total in window, caps, store caveats) via scripts/vendor-insights.py. - Model-in-the-loop faceting: Reads bounded transcripts and writes PII-clean facet JSON capturing goals, outcomes, satisfaction, friction, and session type. - Narrative and HTML rendering: Composes a grounded aggregate narrative with coverage notes and renders a shareable report-<ts>.html file. - Use Case: When asked "what has codex been doing this month", index the codex store, facet the most active sessions, write a narrative with stated denominators, and send the rendered HTML report. ## Quick Start Generate an insights report for the codex agent lane covering the last 14 days and send me the rendered HTML file.

Frequently Asked Questions about vendor-insights

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a usage insights report for codex or copilot sessions?▼

Run scripts/vendor-insights.py index with the vendor name and window, facet selected sessions into JSON files, write a narrative.json, then run the render command to produce a shareable HTML report.

What agent vendors are supported for cross-vendor session insights?▼

Supported vendors are codex, copilot, opencode, antigravity (alias agy), gemini, and claude for parity. Store paths live in the VENDOR_REGISTRY inside the ingest script and must never be hand-resolved.

Does the report include raw transcript content or personal data?▼

No. Facets must be PII-clean with summaries and counts only, no verbatim personal content, secrets, or third-party names. Everything under logs/vendor-insights is gitignored and must never be committed or pasted into chat.

Why does the render step exit with code 2?▼

Exit code 2 means the facet or narrative files are missing for that vendor, so rendering was correctly refused. Complete the facet and narrative steps first; it is sequencing feedback, not an error to suppress.

How should sampling caps be handled when sessions exceed the index limit?▼

Read meta.total_in_window for the true population and treat a capped index as a sample of the newest sessions. State the cap in coverage_notes and never write aggregates as if the sample were the full corpus.