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/bartolli/codanna/graphnpx skills add bartolli/codanna --skill graphgit clone --depth 1 https://github.com/bartolli/codannaWhat 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 | $0.00131 | $0.02271 |
| Opus 5 | $0.00066 | $0.01136 |
| Sonnet 5 | $0.00026 | $0.00454 |
| Haiku 4.5 | $0.00013 | $0.00227 |
Grade A, and why
graph 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 2d 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 — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
What it draws
One HTML file, no server: the codanna index rendered as a pie chart of symbols. Each top-level module owns a wedge sized by its share; inside a wedge symbols fill rings from the centre outwards, best-connected first, so hubs sit near the middle and leaves on the rim. Unlinked symbols form their own inner group. Above the disc a heatmap shows symbols added per day (each symbol dated by the oldest surviving line of its span, via git blame). Hover or click a symbol to see its edge web and the connected symbols; the legend hides (eye) or highlights (label) modules and submodules; the ribbon under the heatmap brushes a date range (drag its handles, use the date fields, or click a year chip) and Refresh replays the codebase growing oldest-first.
Run
node ${CLAUDE_SKILL_DIR}/graph.mjs # whole index, Calls edges, default kinds
node ${CLAUDE_SKILL_DIR}/graph.mjs --root crate::indexing # one module family
node ${CLAUDE_SKILL_DIR}/graph.mjs --relation calls,uses,implements,extends
node ${CLAUDE_SKILL_DIR}/graph.mjs --kinds function,method,struct,trait --dates none --light
- Runs
codanna dumpin the project (--from graph.jsonlreuses a saved dump,--binary PATHpicks the binary) - Output:
.codanna/visualizations/graph-disc-<timestamp>.html, opened in the browser (--no-opento skip,--out FILEto place it) --group <wedge>[/<tint>]-- wedge axismodule(default),language, orkind; tint axismodule(default),kind, orvisibility. The wedge is the angle (share, neighbours adjacent); the tint is the legend tree inside it (three biggest get their own shade, the rest pool).module/kindkeeps the structure and shows what each module is made of;language/modulefor polyglot indexes;module/visibilityfor API surface per module;kind/moduleis a census (edges cross everywhere)--root PREFIXscopes to a module prefix (crate::is dropped,./:://all separate segments); applies to the module segments whatever the wedge axis--kindsdefaults to Function, Method, Struct, Class, Trait, Enum, Interface, TypeAlias, Constant, Macro--relationdefaults tocalls; any of calls, uses, implements, extends, defines, comma-separated--dates blame|git|none--blame(default) dates every symbol by the oldest surviving line of its span (git blame -Mover the working tree; per-file JSON cache at.codanna/visualizations/dates-cache.jsonkeyed by content hash, so warm runs spawn no blame -- delete the file to reset);gituses the file's first-commit day (cheap, but a refactor reads as mass birth);noneleaves the timeline and heatmap empty--unlinked include|drop--dropremoves symbols with no edge over the chosen relations at build time (smaller file, smaller disc); in the page, the legend eye on(unlinked)hides the same set with the cascade animation--name NAMEsets the title (default: project directory name);--lightbuilds the light theme
What ships with it
34 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.
- assets/logo-mask.svg 18 KB
- graph.mjs 9.0 KB runs code
- lib/adapter.mjs 8.4 KB runs code
- lib/dates.mjs 4.8 KB runs code
- lib/dump.mjs 1.9 KB runs code
- lib/vendor.mjs 9.0 KB runs code
- page.css 56 KB
- page.html 13 KB
- page.js 720 KB runs code
- shell.html 5.1 KB
- UPSTREAM.md 5.5 KB
- vendor/graphology.umd.min.js 72 KB runs code
- vendor/hljs/BUILD.md 1.4 KB
- vendor/hljs/c.min.js 4.6 KB runs code
- vendor/hljs/clojure.min.js 4.0 KB runs code
- vendor/hljs/core.min.js 21 KB runs code
- vendor/hljs/cpp.min.js 5.9 KB runs code
- vendor/hljs/csharp.min.js 4.0 KB runs code
- vendor/hljs/github-dark.min.css 1.3 KB
- vendor/hljs/github.min.css 1.3 KB
- vendor/hljs/go.min.js 1.4 KB runs code
- vendor/hljs/java.min.js 2.7 KB runs code
- vendor/hljs/javascript.min.js 6.4 KB runs code
- vendor/hljs/kotlin.min.js 3.3 KB runs code
- vendor/hljs/LICENSE-highlightjs 1.5 KB
- vendor/hljs/lua.min.js 1.9 KB runs code
- vendor/hljs/php.min.js 6.3 KB runs code
- vendor/hljs/python.min.js 3.5 KB runs code
- vendor/hljs/rust.min.js 2.9 KB runs code
- vendor/hljs/swift.min.js 8.0 KB runs code
- vendor/hljs/typescript.min.js 7.6 KB runs code
- vendor/LICENSE-vault-graph 1.0 KB
- vendor/NOTICE.md 5.7 KB
- vendor/sigma.min.js 182 KB runs code
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.
- 2d ago First seen · 91 lines · 131 tokens per session scan A fec5d903a93f
graph is a skill published in the GitHub repository bartolli/codanna (729 stars, last pushed 3d ago), licensed Apache-2.0. It adds 131 tokens to every session and 2,271 once invoked, about $0.0007 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
use-ivygrep
Gather focused local repository context with ivygrep before implementing, debugging, reviewing, or explaining code. Use for coding tasks involving unfamiliar paths, branch changes, stack traces, architectural relationships, callers, dependents, tests, configuration, or documentation. Prefer one bounded context pack…
codemeridian-test-planning
Plan focused tests with CodeMeridian by finding relevant test shields, coverage gaps, impacted behavior, and the smallest useful test set before implementation.
codemeridian-context
Gather minimal, graph-grounded CodeMeridian context before implementation, refactoring, deletion, debugging, or test planning.
codemeridian-frontend
Route frontend HTML/CSS/SCSS work through CodeMeridian's generic frontend-aware tools first, then use cascade and style-duplicate analysis only when the question is truly frontend-specific.
codemeridian-refactor
Plan safer refactors with CodeMeridian by checking graph freshness, exact symbols, impact, tests, duplication, and architecture risk before editing.
human-cognitive-seed
Preserve and strengthen the user's independent reasoning during design, strategy, learning, interpretation, hypothesis work, and consequential decisions, and preserve justified durable context with explicit provenance when supported. Use when a task benefits from human-led judgment, model-building, productive…