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 agentmods add skills/tarekkharsa/agentstack/analyze-usagenpx skills add Tarekkharsa/agentstack --skill analyze-usagegit clone --depth 1 https://github.com/Tarekkharsa/agentstackWhat 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 | $0.00045 | $0.00444 |
| Opus 5 | $0.00023 | $0.00222 |
| Sonnet 5 | $0.00009 | $0.00089 |
| Haiku 4.5 | $0.00005 | $0.00044 |
Grade A, and why
analyze-usage 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 yesterday.
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.
What it actually says
Analyze usage
Use when you want to understand your own footprint: how many tokens you're burning, and which installed capabilities are actually pulling their weight.
Where the data is
Claude Code writes a JSON-lines transcript per session at:
~/.claude/projects/<project-hash>/<session-id>.jsonl
Each line is one event; assistant turns carry token usage, and tool-use events name the tool or skill invoked. It's local and safe to read.
Token burn
Sum token usage across a session or a day. The fields are nested
(message.usage.input_tokens / output_tokens), so use python for anything
beyond a trivial count — don't hand-roll it in a shell one-liner.
Dead-weight detection (the useful one)
- List what's installed:
agentstack more lib list(library skills + servers) and the project's active profile. - Extract which skills/servers were actually invoked, from the transcripts' tool-use / skill-load events.
- The set difference — installed but never invoked — is dead weight: it taxes every session's context window for nothing. Propose pruning it.
Rules
- Read-only and local. Never send transcript content anywhere; report only aggregates — counts, totals, names.
- Best-effort. The transcript format is undocumented and can change — tolerate unknown lines, don't assume a fixed schema.
- Coverage is uneven. This is rich for Claude Code; other CLIs expose less (or nothing). Say which agents you could actually see.
Note
agentstack records some of this natively (activation counts, per-server context
cost). If agentstack more report calls exists in your version, prefer it — it joins
usage to the library for you. This skill is the portable, no-code version.
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.
- yesterday First seen · 51 lines · 45 tokens per session scan A 94757ec11e3b
analyze-usage is a skill published in the GitHub repository Tarekkharsa/agentstack (3 stars, last pushed 18d ago), licensed Apache-2.0. It adds 45 tokens to every session and 444 once invoked, about $0.0002 per session on Opus 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 2026-08-31.
Other skills, from other repositories
lint-js
Lint JS/TS code only. Use before opening a PR when only JavaScript or TypeScript files were changed (no Rust).
feishu
Work with Feishu or Lark bots, docs, sheets, bitables, approval flows, and OpenAPI/MCP setup without hardcoding credentials.
interview
Ask one useful structured question at a time only when material product/implementation choices are genuinely missing; remember answers and produce a brief/spec. Discoverable facts should be investigated instead of asked.
test
Detect the project’s test stack, run the narrowest useful tests, create tests when authorized, and report coverage/gaps honestly.
verify
Exercise the real app/API/CLI and collect observable evidence; tests alone do not count as end-to-end verification.
stack-merge
Bulk-merge a Graphite stack into main via admin-bypass fast-forward push. Replaces the Graphite merge queue for rivet-dev/rivet. Invoke when the user says "merge the stack at ", "ship everything through ", "bulk-merge up to ", "fast-forward main to ", or any request to land multiple stacked PRs at once in one shot.…