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/abilityai/abilities/add-canonnpx skills add Abilityai/abilities --skill add-canongit clone --depth 1 https://github.com/Abilityai/abilitiesWrote 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/abilityai/abilities/add-canon)<a href="https://agentmods.dev/skills/abilityai/abilities/add-canon"><img src="https://agentmods.dev/badge/skills/abilityai/abilities/add-canon.svg" alt="Measured on agentmods" 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 | $0.00211 | $0.08595 |
| Opus 5 | $0.00105 | $0.04297 |
| Sonnet 5 | $0.00042 | $0.01719 |
| Haiku 4.5 | $0.00021 | $0.00860 |
Grade A, and why
add-canon 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 3d 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 — 346 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Add Canon
ℹ️ First, set expectations: before anything else, print one short line with this skill's version and its most recent change — the top entry of
metadata.changelogabove — e.g.add-canon vX.Y — recent: <summary>. Then proceed.
Give an agent a published data layer: a separately-versioned git repository — the canon repo — shared across the fleet, holding each agent's canonical data (the business facts humans and other agents rely on) plus the protocols/ that define inter-agent contracts. Humans and agents co-edit it; each agent is responsible for keeping its own folder true.
The boundary this installs:
| Layer | Where | What lives there |
|---|---|---|
| Working memory | the agent's own repo/workspace | drafts, state, scratch, everything in flight — private |
| Canon | the shared canon repo, agents/<name>/ |
published facts the fleet may depend on — versioned, stamped, owned |
| Protocols | the shared canon repo, protocols/ |
inter-agent contracts (schemas, channels, cadences) — changed via PR |
| Relations | the shared canon repo, agents/<name>/docs/relations/ |
per-counterpart collaboration memory — one doc per counterpart, owned like the rest of the folder |
Relations — the collaboration-memory convention this installs: protocols say how agents agreed to work together; relations record how the work actually went. Each agent keeps one doc per counterpart it collaborates with — docs/relations/<counterpart>.md: working agreements, a capped recent-events log (last 10, older folded into a rolling Earlier: summary), open threads. The standing rule (carried by the CLAUDE.md section): read the relation doc before acting on a counterpart's message; append the outcome before closing the interaction — so an incoming ask is handled as a continuation, not a first contact, and the log updates as a side effect rather than a chore. Each side keeps its own view under own-folder writes; divergence between the pair is a dropped-thread signal, not an error. The docs are ordinary enveloped docs/ files — lint-clean under the two-zone schema with no linter change. Full spec: CONVENTIONS.md § Relations.
Design invariant (do not violate): the canon layer is convention + skills on plain git — no new platform primitive, no new Trinity surface, no sync service. Git supplies versioning, review, audit trail, and human+agent co-editing; CODEOWNERS supplies per-folder review routing. Trinity involvement stays light and optional: the layer is declared in template.yaml (x-canon:) so /discover-agents can see it, and the reconcile schedule rides the normal schedules: machinery. Write scope is own-folder-only: an agent commits directly only inside agents/<its-name>/; anything else — another agent's folder, protocols/, root files — goes out as a branch + PR, never a direct push. Git history is the audit trail, so no extra approval gate sits in front of own-folder writes.
Sibling layers: /add-canon-lint is the law of this layer — run once per fleet against the canon repo, it seeds a deterministic linter (stdlib Python, no LLM) + CI that mechanically enforce the two-zone schema below: every folder keeps facts.yaml (structured claims — the purely lintable zone, one home per key across the fleet) beside enveloped prose in docs/; the linter proves internal consistency on every push, /canon-reconcile verifies external truth on schedule. /add-orchestrator is the routing layer (its /discover-agents scans x-canon: into a canon: field per map node, and /orchestrate serves authoritative-data reads from the canon repo instead of spending a chat turn — writes still route to the owning agent). /add-git-sync is the working-memory durability layer — its hooks manage the agent's own repo, not the canon clone; the clone is gitignored here and synced by the canon skills at use time.
What gets installed into the target agent:
| Artifact | Location | Purpose |
|---|---|---|
.claude/skills/canon-publish/SKILL.md |
agent repo | review + commit own-folder changes; cross-folder → branch + PR |
.claude/skills/canon-consume/SKILL.md |
agent repo | read another agent's published data / a protocol, cited at canon@<sha> |
.claude/skills/canon-reconcile/SKILL.md |
agent repo | scheduled freshness pass over the own folder — verify, stamp, push |
.claude/skills/canon-doctor/SKILL.md |
agent repo | verify the layer end-to-end — credentials, clone, pull, push probe — exact fix per failure |
GH_TOKEN= placeholder |
.env.example |
deployment credential — fine-grained PAT scoped to the canon repo (Step 6b) |
canon/ clone |
agent repo root (gitignored) | working copy of the shared canon repo |
agents/<name>/ (+ seed profile.md, facts.yaml) |
canon repo | this agent's owned folder — its published record, in the two-zone schema |
CONVENTIONS.md, CODEOWNERS, protocols/ |
canon repo (seeded once) | the shared rules of the layer |
x-canon: block |
template.yaml |
declares the layer — repo, folder, write scope, reconcile cadence |
| reconcile schedule | template.yaml schedules: + Trinity MCP |
canon-reconcile, default cron 0 8 * * 1 |
CLAUDE.md ## Canonical Data (Canon) section |
agent repo | wires the skills + states the boundary and the rules |
| fleet enrollment (opt-in, Step 9 — orchestrator context) | member repos + canon repo | same artifacts installed into all/subset of mapped agents; their folders seeded |
What ships with it
8 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.
- 3d ago First seen · 346 lines · 211 tokens per session scan A 3885c51cd7f2
add-canon is a skill published in the GitHub repository Abilityai/abilities (11 stars, last pushed 15d ago), licensed MIT. It adds 211 tokens to every session and 8,595 once invoked, about $0.0011 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
systematic-debugging
Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
auto-perf-optimize
Run agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.
chat-perf
Run chat perf benchmarks and memory leak checks against the local dev build or any published VS Code version. Use when investigating chat rendering regressions, validating perf-sensitive changes to chat UI, or checking for memory leaks in the chat response pipeline.
chat-pet-sprite-creation
Use when creating or changing VS Code chat pet sprite art, sprite sheets, state animations, eye treatments, Stable/Insiders variants, or pet transitions under src/vs/workbench/contrib/chat/browser/widget/media/chatPet.
cpu-profile-analysis
Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…