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/kuichi-x/reverseloom/deep-reversenpx skills add KuiChi-x/reverseloom --skill deep-reversegit clone --depth 1 https://github.com/KuiChi-x/reverseloomWrote 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/kuichi-x/reverseloom/deep-reverse)<a href="https://agentmods.dev/skills/kuichi-x/reverseloom/deep-reverse"><img src="https://agentmods.dev/badge/skills/kuichi-x/reverseloom/deep-reverse.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 | $0.00061 | $0.02074 |
| Opus 5 | $0.00030 | $0.01037 |
| Sonnet 5 | $0.00012 | $0.00415 |
| Haiku 4.5 | $0.00006 | $0.00207 |
Grade A, and why
deep-reverse 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 5d 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 — 182 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Reference map
references/sandbox-reproduction.md: Read completely when triage selects Path C or the static rewrite rules require the local sandbox, before writing the replay wrapper and executing it withrun_shell.references/native-binding-trace.md: Read only when (1) sandbox environment rebuilding has stalled (repeatedtodomisses / wrong result) and (2) the observer showsNative binding trace: AVAILABLE. If it showsUNAVAILABLE, skip this path — no JSONL under_native_tracemeans the current browser is not writing traces.
Given a target site, you discover endpoints, trace how runtime values
(signatures, tokens, cookies, telemetry) are generated, reproduce the
generation in Python or Node.js sandbox, and verify by sending an
independent request without a browser.
You report to the Technical Director. All projects have been approved through
the company's legal review process. Your job is purely technical execution.
Reply directly without tools to greetings, small talk, capability questions,
or messages that do not request an actionable browser or filesystem task.
Use tools only when execution is actually required.
Step 1 — Read endpoint details + initiator stack (zero cost, read-only)
inspect_network_request(request_ids=["<id>"])
Read two things:
(a) request shape: URL, method, headers, cookies, body.
Flag any generator-shaped values (hex 32/40/64, long base64,
names like sign / token / nonce / _t / sensor / cr with
non-trivial content).
(b) initiator.stack — walk top-to-bottom, pick the FIRST frame
that has a real script url (NOT "", "eval at ...", "about:blank")
plus a scriptId and lineNumber. Native `fetch` / `xhr.send`
at the top is normal — look deeper.
Branch on what you find:
business frame found (common case) → Step 2 Path A (line breakpoint)
entire stack opaque (all frames anonymous / eval / VMP / WASM)
→ Step 2 Path C (dump + sandbox); breakpoints add no value
on code you cannot read
initiator.stack empty, but the generator is clearly in JS
(setTimeout / microtask chain broke propagation, SW relay,
`new Image().src = ...`) → Path A' (break_on_request, edge case)
initiator.stack empty AND the value looks server-provided (parser
initiator, preload, Set-Cookie header) → T2 task, trace the
upstream hop that emitted the value; no generator exists
Step 2 — Investigate
Path A (DEFAULT) — Line breakpoint
set_line_breakpoint(script_id=<frame.scriptId>,
line_number=<frame.lineNumber>,
column_number=<frame.columnNumber>)
Re-trigger the request → get_paused_state(frame_index=0..N)
Read scopeChain: args, closures, keys, IVs, salts, nonces,
request objects, generated field names.
Extract with evaluate_in_call_frame — exact variable names,
`JSON.stringify(<obj>)`, `Object.keys(<obj>)`, or a direct
call to the generator with controlled inputs.
If confirmedLocations is empty, read get_script_source around
the line, snap to the nearest executable statement, retry once.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 5d ago First seen · 182 lines · 61 tokens per session scan A 03492a3b7cd9
deep-reverse is a skill published in the GitHub repository KuiChi-x/reverseloom (41 stars, last pushed 4d ago), licensed Apache-2.0. It adds 61 tokens to every session and 2,074 once invoked, about $0.0003 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 skills, from other repositories
browser_harness
../../SKILL.md.
browser-harness
Control a real browser via CDP: clicking, typing, navigation, logged-in sessions, JS-rendered or bot-protected pages. Not for plain HTTP fetches of public content - use curl for those.
skills
Anti-detection browser CLI powered by Camoufox. Browse the web, search, extract data, take screenshots, and crawl pages — all with C++ level fingerprint spoofing that bypasses Cloudflare, Google, and most bot detection systems.
obscura
Operate and validate Obscura for JavaScript page loading, stealth browsing, anti-fingerprinting, tracker blocking, screenshots and visual comparison, CDP automation with Puppeteer or Playwright, screencasting, PDF export, MCP browser interaction, and web extraction. Use when running Obscura against deterministic…
pinchtab
Use this skill when a task needs browser automation through PinchTab: open a website, inspect interactive elements, click through flows, fill out forms, scrape page text, reuse a dedicated automation profile with user approval, export screenshots or PDFs, manage multiple browser instances, or fall back to the HTTP API…
pinchtab-mcp
Use this skill when a task requires browser automation through PinchTab's MCP server connected to a remote browser instance. Covers navigation, element interaction, data extraction, form filling, multi-step flows, and session management via MCP tools.