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/wenyuchiou/research-hub/research-context-compressornpx skills add WenyuChiou/research-hub --skill research-context-compressorgit clone --depth 1 https://github.com/WenyuChiou/research-hubWhat 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 | $0.00074 | $0.01794 |
| Opus 5 | $0.00037 | $0.00897 |
| Sonnet 5 | $0.00015 | $0.00359 |
| Haiku 4.5 | $0.00007 | $0.00179 |
Grade A, and why
research-context-compressor 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 yesterday.
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 — 173 lines — stays where its author put it; the contents beside it link to each section on GitHub.
research-context-compressor
Build a compact, machine-readable workspace memory at the project root
under .research/. Other research-hub skills (project-orienter,
literature-triage-matrix, paper-memory-builder) read these files instead of
rescanning the repo every session, which is where the token savings come from.
This is the foundation skill. Run it once per project and refresh when the project's research question, datasets, or experiment set changes.
When to use
Trigger phrases:
- "Compress this project context for future agents."
- "Create a research manifest so future sessions don't reread everything."
- "Save the key project context before we continue."
- "Build a
.research/folder for this repo."
Not for:
- Compressing source code or generating doc strings — that's a code task.
- Running experiments or analyzing results — those write to
outputs/, not.research/. - Writing the manuscript itself.
Inputs you should read (whichever exist — priority order)
The compressor reads whatever your project has. None of these are required. If a file is missing, that field of the manifest stays empty (see "What NOT to do"). Skim, do not deep-read.
For any project:
README.mdat the repo root — project overview. Single most useful input..research/design_brief.mdif it exists — Stage 3a handoff fromresearch-design-helper(plugin v0.3.12+). Read the frontmatter (project,source,gap_verdict) plus section 1 (Research question) only — informs theproject_manifest.ymlresearch_questionfield. The brief is the authority on the sharpened RQ; the manifest mirrors it. Don't deep-read; skim. If frontmattersource: topic_dossier.gaps.yml#<id>is set, copy that gap-id into the manifest'sprovenance.from_gapfield (forward-compat with the Stage 2 → 3a wire).
For code-based research projects (Python / JS / R / Julia / etc.):
pyproject.toml/package.json/requirements.txt/renv.lock/Project.toml— primary tools.scripts/andnotebooks/— main entrypoints.data/andoutputs/— datasets and artifacts.
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.
- yesterday First seen · 173 lines · 74 tokens per session scan A 947de6d5762a
research-context-compressor is a skill published in the GitHub repository WenyuChiou/research-hub (53 stars, last pushed 1mo ago), licensed MIT. It adds 74 tokens to every session and 1,794 once invoked, about $0.0004 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
code-review-excellence
This skill should be used when the user asks to review a diff or pull request, write review comments, audit code quality, establish review standards, or improve how a team performs code review.
git-workflow
This skill should be used when the user asks to "create git commit", "manage branches", "follow git workflow", "use Conventional Commits", "handle merge conflicts", or asks about git branching strategies, version control best practices, pull request workflows. Provides comprehensive Git workflow guidance for team…
architecture-design
Use only when creating new registrable ML components that require Factory or Registry patterns.
aaai-experiments
Use when designing or auditing AAAI experiments for the broad-AI program committee, including baselines, ablations, statistical significance, robustness, human evaluation, AI-for-Social-Impact and alignment/safety evidence, compute and cost reporting, and reproducibility-checklist alignment for Phase-1 survival.
acmmm-experiments
Use when designing or auditing the experiments of an ACM MM (ACM Multimedia) paper — matched baselines per modality, ablations that isolate the cross-modal fusion, user studies or QoE measurement where the claim is subjective, dataset and media licensing, and honest compute reporting, so evidence supports a multimedia…
ccs-writing-style
Use when revising an ACM CCS paper for a precise threat model, calibrated attack/defense claims, 12-page ACM sigconf compression, double-blind wording, adaptive-evaluation honesty, and the security house style that survives an adversarial SIGSAC program committee.