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/jimezsa/opencolabnpx agentmods add skills/jimezsa/opencolab/deep-researchWrote 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/jimezsa/opencolab/deep-research)<a href="https://agentmods.dev/skills/jimezsa/opencolab/deep-research"><img src="https://agentmods.dev/badge/skills/jimezsa/opencolab/deep-research/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/jimezsa/opencolab/deep-research"><img src="https://agentmods.dev/badge/skills/jimezsa/opencolab/deep-research.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 | $0.00032 | $0.03806 |
| Opus 5.5 | $0.00013 | $0.01522 |
| Sonnet 5.5 | $0.00006 | $0.00761 |
| Haiku 4.5 | $0.00003 | $0.00381 |
Grade B, and why
deep-research scanned grade B with 2 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 1mo 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.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
"script": "git clone https://github.com/jimezsa/papercli.git && cd papercli && make build && sudo install -m 0755 ./bin/papercli /usr/local/bin/papercli", Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl -fsSL https://raw.githubusercontent.com/jimezsa/papercli/main/SKILLS/deep-research/SKILL.md \ How it starts
The opening of the file, as written. The whole thing — 422 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Deep Research Skill
Use this skill for comprehensive scientific research tasks such as state-of-the-art reviews, deep comparisons, research strategy, and evidence-heavy decision support.
If the user later asks an exact follow-up question about a downloaded paper or wants a bounded local verification pass, switch to pageindex-grounded for grounded retrieval over the existing PDF corpus.
Update This Skill
Only do this if the user explicitly asks to update this skill from the GitHub repo.
To refresh this skill directly from the GitHub repo:
curl -fsSL https://raw.githubusercontent.com/jimezsa/papercli/main/SKILLS/deep-research/SKILL.md \
-o SKILLS/deep-research/SKILL.md
Mission
Deliver an institutional-grade findings.md by:
- Running iterative
papercliretrieval across multiple query waves. - Downloading and reading a broad, diverse paper corpus.
- Extracting core ideas, concepts, results, assumptions, and key mathematics.
- Producing a detailed markdown report inside a topic-scoped research run folder, where all claims are grounded by references.
- Producing a companion literature-map block diagram that shows how the main papers or paper families connect.
Prerequisites
papercliis installed and available inPATH.
Non-Negotiable Rules
- Use
paperclias the retrieval backbone. - Read paper content from downloaded PDFs whenever possible.
- Never present uncited factual claims.
- Surface conflicts and uncertainty explicitly.
- Final output must be a detailed markdown file named
findings.mdinside the active research run folder. - Each distinct topic must live in its own dated, topic-slugged folder under
research/. - Maintain
research/INDEX.mdand the run-localRUN.mdmetadata file so later agents can recognize what each research folder contains. - After synthesis, produce a companion literature-map diagram through the shared
block-diagramskill. - The literature map must only show evidence-backed relations such as method lineage, direct comparison, shared benchmark or dataset, critique, or common problem framing.
- Do not invent paper-to-paper influence or citation edges that are not supported by the corpus.
- OpenColab normally provides
OPENCOLAB_PROGRESS_FILEduring provider runs. When it is set, emit bounded JSON progress updates for long-running stages instead of remaining silent until the end.
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.
- 1mo ago First seen · 422 lines · 32 tokens per session scan B d8cef395a933
deep-research is a skill published in the GitHub repository jimezsa/opencolab (11 stars, last pushed 20d ago), licensed MIT. It adds 32 tokens to every session and 3,806 once invoked, about $0.0001 per session on Opus 5.5. A static security scan graded it B with 2 findings (asks for root, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
formulate-hypotheses
Generate testable hypotheses from theory, empirical regularity, anomaly, or explicit explanatory competition. Deductive, inductive, abductive, and competing-hypothesis generation are modes; discriminating predictions and comparison are activated in competing mode.
design-experiment
Translate a falsifiable hypothesis into a statistically defensible experiment. Factorial, ablation, comparison, scaling, and robustness designs are modes, not strategy nodes.
bias-detection
Assess systematic biases in the evidence body — publication bias, reporting bias, and selective outcome reporting. Budget: 40 studies, 40 effect sizes, 40 web searches.
construct-causal-model
Construct and validate an explicit causal model with measurable variables, mechanism edges, evidence links, feedback loops, interventions, counterevidence, and confidence.
map-validity-envelope
Map where a claim, method, or design remains valid by defining variation axes, probing systematic or adversarial/boundary values, detecting breakpoints, and constructing a multidimensional validity envelope. Critical-case selection is an optional inference-maximizing mode.
audit-convergence-independence
Audit claims of independent convergence by tracing shared priors, data, framing, models, prompts, assumptions, or upstream evidence; estimate an effective independent evidence count rather than treating nominal N paths as independent.