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 agents/tajd/projektor/flow-metricsgit clone --depth 1 https://github.com/TAJD/projektorWrote 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/agents/tajd/projektor/flow-metrics)<a href="https://agentmods.dev/agents/tajd/projektor/flow-metrics"><img src="https://agentmods.dev/badge/agents/tajd/projektor/flow-metrics.svg" alt="Measured on agentmods" 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 | $0.00017 | $0.00891 |
| Opus 5 | $0.00009 | $0.00445 |
| Sonnet 5 | $0.00003 | $0.00178 |
| Haiku 4.5 | $0.00002 | $0.00089 |
Grade A, and why
flow-metrics 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 4d 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 — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
get_flow_metrics computes these from indexed transition timestamps for one project.
Pass since/until (epoch seconds) to window it, and granularity: "day" for daily
buckets (default is weekly: the current ISO week plus the preceding five).
The collaboration-shape metrics below measure human attention, not an agent-vs-human split — the question they answer is "where does a human's time go," not "who did the work."
Time-in-state
| Metric | Definition |
|---|---|
leadTime |
ready → done |
cycleTime |
claimed → done |
timeInProgress |
claimed → next stage |
reviewLatency |
in_review → done — the primary human choke point, with a reviewLatencyOverTime trend |
agingWip |
every currently open (in_progress/in_review) issue with its age since claim — a present-state snapshot, not scoped to since/until. Read it against cycleTime's p50/p90 as reference lines: it surfaces stuck items before they finish and skew the percentiles. |
Collaboration shape
| Metric | Definition |
|---|---|
humanInterventions |
human-authored comments plus status bounces out of review, per completed issue |
autonomyRatio |
lease-held time ÷ cycle time (claimed → done), per completed issue |
flowEfficiency |
lease-held time ÷ lead time (ready → done), for issues completed in the window |
autonomyRatio and flowEfficiency look similar but divide by different spans:
autonomyRatio excludes queueing time before an issue was claimed; flowEfficiency
includes it. A project with a long ready→claimed queue will show a lower
flowEfficiency than autonomyRatio even with identical execution.
Volume over time
| Metric | Definition |
|---|---|
wipOverTime |
work-in-progress (claimed but not done) count per bucket |
throughputOverTime |
issues completed per bucket |
cfdOverTime |
cumulative flow diagram: per-bucket counts of issues in each stage (backlogTodo, inProgress, inReview, done). done is cumulative and never decreases; a widening band between two stages is a choke point. |
arrivalVsCompletionOverTime |
issues created vs. completed per bucket, plus net (created − completed) — answers whether the backlog is growing or burning |
bugShareOverTime |
bug share of completed throughput per bucket (total, bugCount, bugSharePercent). Untyped issues count toward total but never bugCount. A rising trend signals more defects shipping, not just more work. |
bugTypeTracked |
whether a task type keyed bug exists in the workspace at all. false means bug share can't be computed (no matching type) — distinct from a genuine 0%. |
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.
- 4d ago First seen · 66 lines · 17 tokens per session scan A 4d8b5524884b
flow-metrics is an agent published in the GitHub repository TAJD/projektor (39 stars, last pushed today), licensed MIT. It adds 17 tokens to every session and 891 once invoked, about $0.0001 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 2026-08-30.
Other agents, from other repositories
retrieval
Two retrieval backends: a lexical index built at compile time that needs no account, and Cloudflare AI Search for hybrid semantic search when you want it.
discovery
How an agent finds these surfaces without being told: the API catalog, DNS-based discovery, an agent-skills index, and the places people actually look.
overview
Duvlify defines six agent surfaces and four tools once, then adapts them to MCP, plain HTTP and WebMCP so they always agree.
mcp
The page outline agents read for line offsets, the in-browser WebMCP bridge and when to run it yourself, the rate limits, and how the surfaces are tested.
builder
Implements a single PLAN.md task in TypeScript (Next.js or Bun). Builds and self-tests; never commits unless told. Dispatched by /build-loop.
reviewer
Read-only code + security gate for a single built task (TypeScript, Next.js or Bun). Returns PASS or FAIL with findings. Dispatched by /build-loop.