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 skills add brightbear2026/research-agent --skill deep-research-workgit clone --depth 1 https://github.com/brightbear2026/research-agentWrote 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/brightbear2026/research-agent/deep-research-work)<a href="https://agentmods.dev/skills/brightbear2026/research-agent/deep-research-work"><img src="https://agentmods.dev/badge/skills/brightbear2026/research-agent/deep-research-work/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/brightbear2026/research-agent/deep-research-work"><img src="https://agentmods.dev/badge/skills/brightbear2026/research-agent/deep-research-work.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00068 | $0.01043 |
| Opus 5 | $0.00034 | $0.00522 |
| Sonnet 5 | $0.00014 | $0.00209 |
| Haiku 4.5 | $0.00007 | $0.00104 |
Grade A, and why
deep-research-work 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 8d 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 — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Deep Research Work
Use the repository's deterministic workflow. Do not reproduce the workflow from memory or create a parallel policy.
Initialize
- Read
CLAUDE.mdfor research integrity and deliverable rules. - Parse
depth=快速|标准|深度(default标准) andmode=regular|plan|execution(defaultregular). - Create a new
projects/<topic-slug>/directory withtools/scaffold.py; never reuse another topic's outputs. - Initialize
tools/workflow_policy.py init --root <project> --mode <mode>. - Check whether the current host exposes the
diagram-designskill. If available, write the selected/saved profile to<project>/.diagram-design(otherwiseprofile: default). If unavailable, use Mermaid as a non-blocking fallback. Do not add a workflow checkpoint for this choice.
Follow the state machine
Run the six phases defined in config/workflow_modes.yaml: kickoff, survey, outline, research, assemble, deliver.
After each phase, run tools/workflow_policy.py advance --root <project>.
- If it returns
needs_confirmation, ask for content confirmation and then runadvance --confirmed. - If it returns
chapters_not_registeredorchapters_incomplete(exit 4), remain in phase four and repair chapter progress. - If it returns
stop, stop. Plan mode must not enter formal research after the outline. - Execution mode contains no repository-defined confirmation point. System permissions, authentication, CAPTCHA, paywalls, and site access controls still apply.
For unavailable sources, try a credible alternative, then record the failure with record-failure. Respect retry budgets and disclose the resulting gap; never loop indefinitely or bypass access controls.
During the research phase, register every outline chapter with register-chapters. Before dispatching a chapter, mark it in_progress; after both the matching draft Markdown and .meta.json exist, mark it completed. On resume, call next-chapter and skip completed chapters. Do not advance to assembly until next_incomplete_chapter is null.
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.
- 8d ago First seen · 60 lines · 68 tokens per session scan A a3f54a6b2591
deep-research-work is a skill published in the GitHub repository brightbear2026/research-agent (2 stars, last pushed 19d ago), licensed MIT. It adds 68 tokens to every session and 1,043 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 2026-08-31.
Other skills, from other repositories
orbit-notion
Open Orbit briefing skill — selected by the Orbit pipeline when Notion is the user's only connected connector, or when the user explicitly scopes their daily digest to Notion. Pulls the past 24 hours of document edits, comments, mentions, and database row changes from the user's authenticated Notion connection and…
instrument-data-to-allotrope
Convert laboratory instrument output files (PDF, CSV, Excel, TXT) to Allotrope Simple Model (ASM) JSON format or flattened 2D CSV. Use this skill when scientists need to standardize instrument data for LIMS systems, data lakes, or downstream analysis. Supports auto-detection of instrument types. Outputs include full…
baoyu-youtube-transcript
A tool for downloading the written captions, subtitles, chapter information, speaker labels, and cover image from a YouTube video using its URL or ID.
feishu
Work with Feishu or Lark bots, docs, sheets, bitables, approval flows, and OpenAPI/MCP setup without hardcoding credentials.
read
Reads URLs and PDFs by fetching source content, defaulting to concise summaries for plain read requests and clean Markdown when asked to convert, save, quote, cite, or feed downstream work. Use when users ask in any language to read, fetch, check, summarize, quote, cite, convert, or save a URL or PDF. Not for local…
overleaf-sync
A two-way connection between a local paper folder and Overleaf, a web-based LaTeX editor for writing research papers. It lets you move changes between the local files and the shared Overleaf project.