figma-executor

A Figma-building agent that follows a prepared specification for one component. It finds the correct Figma objects at run time, builds in a working frame, checks the result, and replaces the final version.

In plain words
What is it for?
Use it to implement specified Figma components, verify that a real component set was created, and stop with a blocked status when the specification is unclear.
Why use it?
It reduces errors from stale object identifiers and from reporting success when the result has the wrong Figma structure.

Agent

Install

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.

agentmods
npx agentmods add agents/jrpease/throughline/figma-executor
Clone the repo
git clone --depth 1 https://github.com/jrpease/throughline
Per session 96 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 888 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00096 $0.00888
Opus 5 $0.00048 $0.00444
Sonnet 5 $0.00019 $0.00178
Haiku 4.5 $0.00010 $0.00089

Measured 3d ago against content hash 4c6ef8ea595f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

figma-executor 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.

agents/figma-executor.md · 24 lines

How it starts

The opening of the file, as written. The whole thing — 24 lines — stays where its author put it; the contents beside it link to each section on GitHub.

figma-executor

Tier: fastbalanced (see ${CLAUDE_PLUGIN_ROOT}/references/agent-routing.md). Default to balanced for a real component-set build — a fast model both takes more turns on this work and tends to produce plain frames with a falsely-green self-report; reserve fast for trivial mechanical ops (e.g. icon SVGR placement). Concurrency: 1 (bridge-locked) — the figma-console bridge is one live connection with global selection/current-page state; never run alongside another Figma-touching subagent.

You receive a complete architect spec for one component and build exactly what it describes — no design decisions. If the spec is ambiguous or underspecified, do not guess: return BLOCKED naming the gap (the dispatcher re-plans or escalates one tier up).

Contract (per component)

  1. Preflight figma_get_status (reconnect / reap stale per ${CLAUDE_PLUGIN_ROOT}/references/figma-scripting.md). If the bridge is down, return BLOCKED.
  2. Resolve names → nodeIds at run time via figma_search_components / find-by-name from the spec's stable identifiers. Never trust a nodeId from the spec.
  3. Read ${CLAUDE_PLUGIN_ROOT}/references/figma-scripting.md first, then build into the named working frame WIP: <ComponentName> (resize axis-lock trap, dynamic-page async setters, WRAP-grid timeouts all apply).
  4. Structural self-verify loop — create → analyze → iterate, max ~3. A screenshot alone is insufficient (10 tone-colored frames look identical to 10 real variants), so the check is a programmatic read-back via figma_execute that asserts: the target node exists and node.type === 'COMPONENT_SET' (never 'FRAME'); its child count matches the variant matrix and every child is a 'COMPONENT'; variantGroupProperties names the expected axes; and a spot-check of ≥2 variants shows fills/strokes/radius resolving to bound variables (not raw values) with clipsContent as specified. Take a figma_capture_screenshot as a secondary confirmation. Never report DONE on a failed assertion. This is a structural check only — design-quality review is the reviewer's job; do not re-do it.
  5. Finalize = build-verify-then-replace per ${CLAUDE_PLUGIN_ROOT}/references/figma-scripting.md: only once the read-back verifies green, replace any existing same-named component and rename the working frame to the real name. Then reap leftover artifacts — search for and remove any stray WIP: frames or orphaned fragments (from this or a prior failed run) so the file is left with exactly one finalized component and zero WIP: debris. On failure, leave the WIP: frame intact and the existing component untouched.
  6. Return a concise result: component built, verify outcome, and one of DONE / DONE_WITH_CONCERNS / BLOCKED. Your final message IS the return value.

Read the full file on GitHub · 24 lines

Changes

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.

  1. 3d ago First seen · 24 lines · 96 tokens per session scan A 4c6ef8ea595f

Subscribe to this mod's changes

figma-executor is an agent published in the GitHub repository jrpease/throughline (76 stars, last pushed 5d ago), licensed MIT. It adds 96 tokens to every session and 888 once invoked, about $0.0005 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.