Skip to content
DeepWaterIMRPublic

About

Agentic AI tools and knowledge packs for fisheries science

Resources

Contributing

Security policy

Stars

0 stars

Watchers

1 watching

Forks

Repository files navigation

Fiskwise — Agentic AI tools for fisheries science

Version 0.2.0 (2026-11-04)

Publish Fiskwise License: MIT

Explore the Fiskwise website

For coding agents receiving an install/update/uninstall request: read AGENTS.md, then execute its matching full lifecycle skill. A first-time install https://github.com/DeepWaterIMR/fiskwise bootstraps the install skill directly from this repository. The user does not need any additional prompt text.

Fiskwise brings together AI skills, scientific knowledge, and R tools to help researchers conduct fisheries science with agents such as Claude Code and Codex. Its knowledge packs guide data preparation, analysis, modelling, scientific writing, and presentation, with researchers directing the work and evaluating the results.

The ecosystem builds on R, tidyverse, and Quarto. The name combines FISK — Fisheries Intelligence Skill Kit with wise, and draws on the Scandinavian word fisk. Developed at the Institute of Marine Research, Norway, its components can be used independently or together. Raw confidential data must remain local.

AI knowledge packs

Pack Purpose Documentation
wise-writing Scientific writing, reproducible documents, and review Repository
wise-presentations Scientific presentation planning, creation, and visual review Learning hub
BAIT Access, preparation, analysis, and visualization of IMR Biotic data and Norwegian catch data Documentation
index-wise Development, validation, and reporting of survey indices using sdmTMB Repository
spict-wise SPiCT stock assessments, diagnostics, sensitivity analyses, and advice Repository
gadget-wise gadget3 fisheries and ecosystem models: structure, data, fitting, diagnostics, projections, and reports Repository

R packages

Package Purpose
ggOceanMaps Ocean maps with bathymetry
ggFishPlots Fish life-history relationships
ggSurveyPlots — planned Fisheries survey visualization

Start with your agent

install https://github.com/DeepWaterIMR/fiskwise

Later, simply say update Fiskwise or uninstall Fiskwise. The agent loads the matching workflow; no scripts or extra instructions are needed from you.

Use a local coding agent with repository and file/shell access: Claude Code, Codex, Antigravity, Mistral Vibe Code, Gemini CLI, Cursor, Copilot CLI, or OpenCode. The installer handles their documented skill-discovery locations. Browser-only chat and hosts that cannot read/write local files cannot install local skills.

Core installation registers wise-writing first, even when R is missing. Fiskwise then guides toolchain checks and asks for revocable permission before installing/updating managed R packages (including ggOceanMaps and ggFishPlots). R, Quarto CLI and system tools require separate approval. Declining package permission keeps the skills usable and reports incomplete runtime setup. Presentations and method packs are installed on demand; BAIT is offered to IMR users with its own privacy/database choices.

Use Update Fiskwise for installed software, or Uninstall Fiskwise for an ownership-aware removal preview. Removal always requires confirmation; existing/adopted installations and databases are preserved by default. Database refresh and removal are separate choices. See Getting started and the lifecycle implementation.

Fiskwise owns fiskwise-workflow, fiskwise-install, fiskwise-update and fiskwise-uninstall, plus fiskwise-check for readiness and repair. wise-writing owns shared r-code-style, r-package-setup and quarto-reporting; specialist methods stay in their packs. Fiskwise research uses R, tidyverse and Quarto, with r-code-style for new/edited R code. Website/maintenance work retains suitable tooling. Delegate only when total expected cost is lower, including coordination and review. Packs remain independently usable.

Existing skill IDs remain stable; see the migration guide. Cite the underlying methods and software using each component's own citation instructions. This hub has no scholarly DOI.

Website and shared theme

The hub uses hand-authored HTML/CSS, with no JavaScript framework or Quarto dependency. The build uses only Python's standard library:

python3 scripts/build_site.py
python3 scripts/check_site.py
python3 -m unittest discover -s tests
python3 -m http.server 8080 --directory dist

GitHub Actions builds and publishes dist/ to GitHub Pages. packages.json is the component inventory. BRAND.md documents the shared SVG assets, design tokens, and synchronization with existing pack websites. Scientific document and slide themes are maintained independently.

Optional responsive and keyboard browser checks use Playwright: run npm ci, npx playwright install chromium, then npm run test:browser with the preview server running. The site build and deployment do not require Node.js or Playwright.

Contributing · Migration · Brand assets · Migration ledger · Hub license

Ecosystem versioning

Every ordinary commit increments patch by 0.0.1 and records the update date. Minor (+0.1) and major (+1) versions change only at the user's explicit request. Knowledge packs use VERSION; R packages use DESCRIPTION Version/Date. See the shared version policy.

About

Agentic AI tools and knowledge packs for fisheries science

Resources

Contributing

Security policy

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages