Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/mishahanin/heading-osnpx agentmods add skills/mishahanin/heading-os/deep-research-advanceWrote 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/mishahanin/heading-os/deep-research-advance)<a href="https://agentmods.dev/skills/mishahanin/heading-os/deep-research-advance"><img src="https://agentmods.dev/badge/skills/mishahanin/heading-os/deep-research-advance/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/mishahanin/heading-os/deep-research-advance"><img src="https://agentmods.dev/badge/skills/mishahanin/heading-os/deep-research-advance.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 187 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
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 5.1 | $0.00159 | $0.02549 |
| Opus 5 | $0.00079 | $0.01274 |
| Sonnet 5 | $0.00032 | $0.00510 |
| Haiku 4.5 | $0.00016 | $0.00255 |
Grade A, and why
deep-research-advance 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 7d 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 — 244 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Deep research advance
Multi-source, verified deep research. The headless script handles token-heavy acquisition (Perplexity) and reasoning/verification (Kimi). Claude takes the intermediate output, audits it, writes a cited report, and proposes an Odin distillate for human approval before anything reaches the brain.
Guardrail (non-negotiable, state first)
This skill sends the user's question and the gathered web corpus to third-party clouds: Perplexity (acquisition), Kimi (reasoning and verification). Both are outside the 31C data boundary.
- NEVER run on a private or internal topic. CRM contacts, pipeline data, Odin
brain content, Exchange mail, Telegram history, partner names, pricing -- none
of these belong in the question text or in
--domains. - If the request references private data, refuse immediately. Point to
/recall(workspace-wide semantic memory) or/odin(brain-scoped reasoning) instead. - If in doubt whether a topic is public-web, ask before running.
Phase 0 -- Scope
If the question is underspecified -- no clear domain, no time horizon, no geography where those matter -- ask 2-3 focused clarifying questions before running. Examples worth asking: "Which jurisdiction's regulation?", "What time window?", "Vendor landscape or technical depth?" Do not pad; ask only what materially changes the research angle.
Confirm the topic is public-web before proceeding.
Phase 1 and 2 -- Run the headless script
Run the acquisition and reasoning phases:
python scripts/deep-research-advance.py "<question>" [--depth N] [--critical] [--domains a.com,b.com]
Notes on flags:
--depth Nsets the number of Perplexity angle queries (default 4, max 8).--criticalforces the adversarial audit governor in Phase 3 regardless of source count.--recency {hour|day|week|month|year}sets the Perplexity time window. Default is none (full index) — correct for evergreen/footprint research. Use a window only for genuinely recent-events research; a narrow window starves a footprint search and returns off-topic recent noise.--domainsaccepts a comma-separated allow-list of domains to constrain Perplexity queries.--audiois a skill-level flag handled in Phase 5. Do NOT pass it to the script.
What ships with it
1 file 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.
- 7d ago Changed e3d586193b71
- 11d ago First seen · 244 lines · 159 tokens per session scan A 5c6b0c130817
deep-research-advance is a skill published in the GitHub repository mishahanin/heading-os (11 stars, last pushed 2d ago), licensed Apache-2.0. It adds 159 tokens to every session and 2,549 once invoked, about $0.0008 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
shogun-bloom-config
An interactive wizard that creates model-routing settings from your AI subscription choices. Model routing decides which available model handles each kind of task.
shogun-screenshot
A screenshot tool for getting images from a computer or web page and then cropping, resizing, or masking sensitive information. Playwright is a browser-automation tool used here to capture web pages.
shogun-model-list
A reference table of AI command-line tools, their available models, subscription requirements, and maximum Bloom capability levels. Bloom's Taxonomy is a scale describing thinking tasks, from remembering information to creating new designs.
shogun-model-switch
A live-switching tool for changing which command-line AI agent, model, and reasoning mode is running. It updates settings, restarts the agent, and refreshes the displayed session information.
shogun-readme-sync
A skill that compares an English README with its Japanese counterpart and brings them back into alignment. A README is the project document that explains what the software is and how to use it.
shogun-agent-status
A skill that shows whether the agents in a multi-agent Japanese feudal-themed team are working, waiting, missing, or assigned tasks. It combines terminal session status, task files, and unread messages.