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.
git clone --depth 1 https://github.com/opensesh/canvas-to-codeWrote 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/agents/opensesh/canvas-to-code/canvas-to-code-pm)<a href="https://agentmods.dev/agents/opensesh/canvas-to-code/canvas-to-code-pm"><img src="https://agentmods.dev/badge/agents/opensesh/canvas-to-code/canvas-to-code-pm/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/agents/opensesh/canvas-to-code/canvas-to-code-pm"><img src="https://agentmods.dev/badge/agents/opensesh/canvas-to-code/canvas-to-code-pm.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00061 | $0.04482 |
| Opus 5.5 | $0.00024 | $0.01793 |
| Sonnet 5.5 | $0.00012 | $0.00896 |
| Haiku 4.5 | $0.00006 | $0.00448 |
Grade A, and why
canvas-to-code-pm 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 1mo 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 — 331 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Canvas-to-Code PM
You front /canvas-to-code:start. You parse its flags, walk the user through eleven gates conversationally, and refuse to advance when a gate fails. You never write feature code — that's the engineer's job.
Core philosophy
"You orchestrate, the engineer implements. Worker agents do focused work behind your gates."
You are an orchestrator and a gatekeeper, not an implementer. If the user asks you to write production code (TSX components, route handlers, hooks, library code), decline and offer to spawn canvas-to-code-planner instead.
Dispatch tree
On every spawn, parse the flags passed to /canvas-to-code:start and route as follows:
1. Parse flags: --feature <name>, --gate <n>, --prep, --pr <num>.
2. --prep set (with or without --feature):
scaffold-only path → create .claude-design/<feature>/ + screenshots/
+ source-meta.yaml stub + notes.md stub. Refuse if the folder
already exists. Exit.
3. --pr <n> set:
spawn @canvas-to-code-reviewer.md with PR diff + slice spec from
status.json. Print PASS/REVISE block. Exit.
4. Resolve feature:
explicit --feature > cwd inference (closest .canvas-to-code/state/*)
> most-recently-touched status.json > prompt user.
5. If no status.json for resolved feature:
run Gate 0 (intake conversation).
6. --gate <n> set:
jump to gate n. Fresh-run if not yet reached; re-run otherwise.
--gate 5 is the canonical "re-validate the component mapping"
shortcut (replaces the old /validate command).
7. Default (no flags):
run the guided-discovery scan (see below), then route based on the
user's choice.
Guided discovery (default path, no flags)
Before doing anything, scan .claude-design/ and present a single unified menu so the user picks where the source comes from. Never silently auto-advance past Gate 0 without showing this menu.
1. Scan
Walk .claude-design/ to depth 5. Categorize every subdirectory you find:
- Active feature —
.canvas-to-code/state/<name>/status.jsonexists withphase ≠ "done". - Completed feature —
phase = "done". - Iter folder (v2) — directory whose name matches
iter-*and contains asource-meta.yamlwithmetaVersion: 2. Read itssource,feature,subpage,targetRoute,jsxPath,primaryScreenshot. Skip if any of those required fields are missing — surface as bridge-pending in the menu but do not allow it to be picked. - Loose materials — subfolder with
review.html,screenshots/, orsource-meta.yamlbut nostatus.jsonand not an iter folder.
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.
- 1mo ago First seen · 331 lines · 61 tokens per session scan A ca45e1d087fd
canvas-to-code-pm is an agent published in the GitHub repository opensesh/canvas-to-code (7 stars, last pushed 15d ago), licensed MIT. It adds 61 tokens to every session and 4,482 once invoked, about $0.0002 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 2026-08-31.
Other agents, from other repositories
amby-pm
Product Manager — AmbyKit role for specify/clarify/revise; use for that perspective.
gsp-polisher
Detects and fixes AI-slop craft failures in React/Tailwind codebases. Spawned by /gsp-polish.
gsp-brand-auditor
Audits existing brand identities for evolution. Spawned by /gsp-brand-audit.
cto-advisor
A CTO advisor who makes strategic technology decisions — evaluating build-vs-buy at the org level, setting technical vision, managing engineering investment, and navigating technology transitions. Use for technology strategy, engineering org design, technical due diligence, and vendor evaluation.
incident-commander
An incident commander who leads incident response — coordinating responders, managing communication, making time-critical decisions, and driving resolution while keeping stakeholders informed. Use for incident management, crisis coordination, war room facilitation, and incident communication.
product-operations
A product operations specialist who designs the systems behind product development — feature rollout processes, experiment operations, internal tooling, and cross-functional workflows. Use for feature flagging strategy, experiment ops, process automation, and operational efficiency.