checking-member-access
Explains what a member or a role can do in a PostHog project, using the access control MCP tools. Use when the user asks what someone can see or edit, who can…
Validate and publish a canvas source project safely: the source-project shape, declared capabilities, reading the current version pointer, iterating on validation diagnostics, guarded publishing with expected_current_version_id, staging a draft build and promoting it, waiting
$ npx -y skills add posthog/posthog --skill validating-and-publishing-canvases --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/validating-and-publishing-canvasesContext preview
The summary Claude sees to decide when to auto-load this skill.
Validate and publish a canvas source project safely: the source-project shape, declared capabilities, reading the current version pointer, iterating on validation diagnostics, guarded publishing with expected_current_version_id, staging a draft build and promoting it, waiting
name: validating-and-publishing-canvases description: > Validate and publish a canvas source project safely: the source-project shape, declared capabilities, reading the current version pointer, iterating on validation diagnostics, guarded publishing with expected_current_version_id, staging a draft build and promoting it, waiting out the queued build, and recovering from a 409 version_conflict or a 429 capacity limit without overwriting concurrent work. Use whenever a canvas edit is ready to save, a draft build is wanted, a canvas publish or build returns diagnostics or a conflict, or a task needs to understand canvas version history.
A canvas's source lives in PostHog, versioned per publish. Publishing is guarded: every edit is based on a specific version, and the server refuses to overwrite newer work. Every publish queues a server-side build, and the canvas renders the last successful build.
`canvas-source-retrieve` returns:
`dependencies` (exact platform-pinned versions), `canvasSdkVersion`, `capabilities`.
It is `null` for a canvas that has never been published — pass that `null` on the first publish.
Keep `index.html` and `dependencies` exactly as returned. You may add relative source files and admitted assets to the project. Use `?worker` for a self-contained module worker and represent binary assets as base64 entries in `assets`; new npm dependencies or dependency-version drift fail validation.
The host enforces `project.capabilities` at runtime, so an undeclared `ph` call builds fine and then dies in the rendered canvas. Declare:
canvas reads through `ph.connectors.call`, listing every tool it calls on that provider. A provider is a native id (`github`) or `mcp:<server host>` (`mcp:mcp.calendly.com`). Unknown providers, unregistered native tools, and private MCP hosts fail validation; every declared tool must have `is_read_only: true` in the catalog. An upstream hint alone does not grant access. A canvas with connectors cannot declare shared state.
external stylesheet, image, font, media file, or frame. Remote scripts remain blocked. Do not include paths, credentials, queries, fragments, or wildcards. The host must be public: loopback and private IPs, single-label names like `intranet`, and the `.local`, `.localhost`, `.internal`, and `.home.arpa` suffixes are all rejected, so a local dev host such as `https://localhost:8010` fails validation with an `invalid_network_origin` error. Data sent to a declared origin leaves PostHog and appears in the capability review before promotion.
Before validation, inventory every literal external URL in every source file. Classify navigation links and `ph.openExternal()` URLs as navigation; they do not need a network origin. For every request or resource URL, declare its scheme + host + optional port only. Include every origin a request redirects to and every secondary origin a stylesheet references for fonts or images. Never infer that one CDN hostname covers another.
Validation rejects undeclared literal calls and resource URLs (`capability_missing_*` diagnostics) so you can fix them before publishing. Dynamic URLs and redirect destinations cannot be inferred, so the inventory is still required even when validation is clean.
`canvas-validate-create` is side-effect free; call it as often as needed. Diagnostics carry `severity`, a stable `code`, a `message`, and (for file-specific problems) `path` and `line`:
(bare imports are limited to the dependencies returned in the source project), `forbidden_dynamic_import` / `forbidden_require` / `forbidden_inline_script`, `invalid_path`, `capability_missing_insight` / `capability_missing_capture_event` / `capability_missing_inline_queries` / `capability_missing_agent_requests` / `capability_missing_network_origin`, `dependency_not_admitted` / `dependency_version_mismatch`, `platform_token_redeclared` (a CSS variable named like a Quill token that the platform stylesheet sets on every element, so the value never applies; prefix your own variables), and path/size violations.
reaches for the network directly. Declare the exact HTTPS origin or use the `ph` bridge.
Publishing goes live immediately and is the default way to save a change, for a canvas's first version and for every follow-up edit. Every version records who published it and which task did the work, so the history stays reviewable after the fact. Stage a draft instead only when the user asked for a draft, a preview, or a review step — see "Draft, then promote" below.
Two ways to publish, both guarded:
file's complete content, or deletes it with `content: null`). Prefer this for small changes to a large project; the guard is mandatory here because a diff's meaning depends on its base.
For a w
:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Repo: posthog/posthog
Explains what a member or a role can do in a PostHog project, using the access control MCP tools. Use when the user asks what someone can see or edit, who can…
Analyze the most expensive users in AI observability and explain why they cost so much. Use when the user asks about top spenders, expensive users, per-user…
Author continuously-running online evaluations in PostHog AI observability, grounded in real failure modes you've identified. Use when the user wants…
Find where an AI/LLM application is failing in production and surface the failure patterns, working from real traces. Use when someone wants to understand…
Investigate AI observability clusters — understand usage patterns in AI/LLM traffic, compare cluster behavior, compute cost/latency metrics, and drill into…
Investigate LLM spend in PostHog — total cost over time, cost by model, provider, user, trace, or custom dimension, token and cache-hit economics, and cost…