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/cogni-work/insight-wave/source-fetchergit clone --depth 1 https://github.com/cogni-work/insight-waveWhat 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 | $0.00233 | $0.04356 |
| Opus 5 | $0.00117 | $0.02178 |
| Sonnet 5 | $0.00047 | $0.00871 |
| Haiku 4.5 | $0.00023 | $0.00436 |
Grade A, and why
source-fetcher 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 3d 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 — 255 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Source Fetcher Agent (inverted pipeline, Phase 3 — cobrowse-only)
Role
You run in one of two modes, set by MODE (default recover):
recover(default — the original job): you take a list of WebFetch-miss URLs (the orchestrator already tried WebFetch in Phase 2 and recorded themunavailable+cobrowse_eligible), recover each via theclaude-in-chromebrowser extension, and record availability. Successful recoveries go to the shared<knowledge-root>/.cogni-knowledge/fetch-cache/(URL-keyed), overwriting the curator's negative-cache entry. Failures get a cobrowse negative-cache entry so a re-run within the freshness window does not re-attempt.topup(additive enrichment): you take a list of URLs that already have a usableokbody but are thin primary-tier sources (the curator flagged themcobrowse_topup_eligible). WebFetch returns a capped extract; the browser reads the fuller rendered text. You cobrowse each and supersede the cached body only when the browser body is strictly longer than the givenBASELINE_WORDS— otherwise you leave the existing body untouched. A top-up never writes a negative-cache entry and never degrades a good source: a failed or not-longer top-up is a silent no-op on the cache.
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.
- 3d ago First seen · 255 lines · 233 tokens per session scan A 7379e1efcdac
source-fetcher is an agent published in the GitHub repository cogni-work/insight-wave (12 stars, last pushed 3d ago), licensed Apache-2.0. It adds 233 tokens to every session and 4,356 once invoked, about $0.0012 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
alchemist
Creative technologist who sees the browser as an unexplored physics engine. Consult when building UI that needs to feel alive - scroll-driven reveals, morphing transitions, spatial animation systems, anything where the interaction itself IS the product. Thinks in weight, tension, and breath before thinking in code.…
praman-sap-planner-cli
SAP UI5 test planner via Playwright CLI. Token-efficient alternative to MCP planner. Generates test plan + gold-standard spec using CLI commands.
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.
FAI Browser Agent
Browser automation agent — navigates websites, extracts data, and executes web workflows using Playwright MCP and vision analysis. Domain-restricted, no credential entry, human approval for transactions.
dogfood-persona
Drives the running whiteboard app as a realistic end-user persona via the Playwright MCP browser tools, to surface friction (bugs, missing affordances, confusing/slow/dead-end flows). Spawned by the dogfood-triage workflow and by the review workflow's optional Dogfood phase. Pass the persona, goal, app URL, and (for…
e2e-tester
Use for end-to-end and smoke testing of critical user paths across viewports. Pairs with a browser-automation MCP (for example Playwright) when one is available.