Borrowing it
Nothing to install: this file belongs to Blushyes/reef. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/Blushyes/reef/main/.agents/skills/developing-reef/SKILL.mdgit clone --depth 1 https://github.com/Blushyes/reefWrote 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/blushyes/reef/developing-reef)<a href="https://agentmods.dev/skills/blushyes/reef/developing-reef"><img src="https://agentmods.dev/badge/skills/blushyes/reef/developing-reef/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/blushyes/reef/developing-reef"><img src="https://agentmods.dev/badge/skills/blushyes/reef/developing-reef.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00198 | $0.03845 |
| Opus 5.5 | $0.00079 | $0.01538 |
| Sonnet 5.5 | $0.00040 | $0.00769 |
| Haiku 4.5 | $0.00020 | $0.00384 |
Grade A, and why
developing-reef 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 today.
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 — 202 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Developing Reef
Use this skill as the project onboarding guide for non-test Reef changes. The workspace contains a renderer-neutral Rust app engine plus the Reef TUI frontend. Every host must stay responsive, and expensive work must not run from a renderer.
Core Architecture Rules
- Keep
ui::*::renderon cached state only. Do not call git, filesystem walks, diff generation, syntax highlighting, external processes, or long formatting from render. Renderer-specific media geometry may be cached during render; decoder open/rebuild work starts from the host tick. - Treat input handlers as intent dispatchers. They decode terminal input and dispatch
reef_app::AppCommand; they must not directly own business state or do blocking host work. - Route expensive work through
reef-app's task coordinator; merge worker results fromReefApp::step. - Put UI-independent logic in
crates/reef-core; shared host filesystem services such as the unified preference store belong incrates/reef-io; keep ratatui/crossterm rendering and input orchestration incrates/reef-tui. - Put renderer-neutral app state, async scheduling, worker-result merge, settings state, nav/history, preview/search/git/graph orchestration in
crates/reef-app. - Renderer hosts dispatch preview definition requests with a typed source cursor and viewport budget. For local backends,
reef-appmarks the initial workspace index stale and schedules its build fromstep; remote backends keep that unsupported work idle.reef-appowns enrichment- and workspace-index-bound retry, definition/reference resolution, candidate state, selected-candidate preview, history, and committed jumps; hosts own only coordinate mapping and popup geometry. - Navigation locations retain the exact source line and UTF-8 byte range for jumps and highlights. Their bounded display snippets are cropped around that target range so every renderer can keep the matched identifier visible without reading candidate files or reconstructing source lines.
- Location history is renderer-neutral and stores the active surface, path, UTF-8 cursor, and vertical/horizontal scroll. Hosts sample their current native position only when dispatching an explicit navigation and pass it to the shared back/forward commands; ordinary cursor movement and scrolling do not create history entries.
- Graph Changed-files selection is renderer-neutral. Hosts dispatch file-navigation commands;
reef-appfollows the visible list/tree order, skips collapsed descendants, updates selection immediately, and schedules the matching commit diff without moving focus away from the Changed-files panel. - Hosts construct
ReefAppfromAppConfig;AppStateand its construction details stay private toreef-app. - Keep terminal-only state in
crates/reef-tui: ratatui layout caches, hit-test registry, terminal image protocol, text-selection geometry, mouse row/column mapping, scroll pacing, leader/chord timers, popup rects, and the live TUI theme object. - Keep inline-video decode and terminal protocol encoding off the TUI loop. Bound decoded frame dimensions and playback FPS by terminal wire cost before starting ffmpeg; playback ticks may drain decoded frames, submit only the newest due frame to a bounded encoder, and merge completed protocol payloads without waiting.
- Prefer stale cached UI over blocking. Show old data plus loading/stale/error status instead of waiting during tab switches or hover/mouse movement.
- Use generation tokens for async results. Late results from older requests must not overwrite newer selections or newer snapshots.
- Keep each tab/panel independently refreshable. Adding a feature should not require another tab to render before data can update.
- When a renderer needs preview content beyond the bounded display projection, expose that content through a typed, renderer-neutral
PreviewBodySnapshot; do not make a host reparse source content while selecting or rendering a preview. - Preserve complete structured source separately from its bounded visible-line projection. Reef must not identify or name third-party formats; renderer-specific recognition belongs to the renderer that consumes this generic source.
What ships with it
3 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.
- today Changed · +6 lines c3f94708b443
- 14d ago Changed · +45 lines 9daa21b47835
- 1mo ago First seen · 151 lines · 198 tokens per session scan A c21ef1219440
developing-reef is a skill published in the GitHub repository Blushyes/reef (84 stars, last pushed yesterday), licensed Apache-2.0. It adds 198 tokens to every session and 3,845 once invoked, about $0.0008 per session on Opus 5.5. 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
refocusing
Use when a long-running agent may have lost the product outcome, when context was compacted, when work has crossed a major boundary, or when a fresh path-based trace is needed before the next consequential action. Not for fresh-session orientation, waking an idle seat, or checkpoints.
rulesync
Generates and syncs AI rule configuration files (.cursorrules, CLAUDE.md, copilot-instructions.md) across 20+ coding tools from a single source. Use when syncing AI rules, running rulesync commands, importing or generating rule files, or managing shared AI coding configurations.
establishing-project-context
Use when the user asks to establish shared project language, or project work exposes a conflicting, renamed, or deprecated domain term that needs active semantic modeling. Routine small tasks stay on the fast path.
gitnexus
A code-graph analysis add-on for examining an existing codebase, including symbols, call paths, execution flows, and effects across repositories. It can query GitNexus through its command-line or MCP interfaces.
energy-procurement
Codified expertise for electricity and gas procurement, tariff optimization, demand charge management, renewable PPA evaluation, and multi-facility energy cost management. Informed by energy procurement managers with 15+ years experience at large commercial and industrial consumers. Includes market structure analysis…
loop
Full execution protocol for MODE: LOOP — the compound-engineering loop: brainstorm → plan → build → review → improve, iterating under defense-in-depth stop conditions with generator/critic separation, durable resumable state, and mandatory compounding learning capture. Loaded on demand by the architect when the loop…