← 2.1.16 Primitive inconclusive · runtime-test

Combo Test: Skill-Driven Task Templates

Skills can define task graph templates that get instantiated on invocation.

Skills can define task graph templates by instructing Claude to invoke TaskCreate and TaskUpdate for each task in a workflow.

How It Works

A skill file contains instructions that specify a multi-task workflow structure, including task names, arguments, and dependency constraints. When invoked, Claude interprets the skill, calls TaskCreate for each task, then uses TaskUpdate to establish ordering constraints. The skill itself is the template: no separate templating system exists. Refinement happens via hot-reload (updating skill instructions), not via template versioning.

What the Test Found

The combo test created a skill defining a 3-task sequence (Setup, Work, Cleanup) with linear dependencies. Invocation worked: Claude read the instructions and created the task graph using available tools. The test marked this "VALIDATED (by design)," but this reflects confirmation that Claude follows instructions, not validation of a robust templating mechanism. The test acknowledged the actual primitives: instructions → tool calls → realized graph.

Why It Matters

This pattern enables workflows to live as declarative text (in skill definitions), reducing hard-coded graph construction in code. Hot-reload allows on-the-fly refinement without redeploying task runners. It folds templating cost into skill authoring, not runtime infrastructure.

Caveats

  • No strong parameterization: argument interpolation depends on Claude's text parsing; no schema-validated slots
  • Unpredictable task IDs: returned by TaskCreate based on session state; templates cannot reliably capture identities for later reference
  • Hot-reload scope: updates skill text but does not mutate already-instantiated tasks
  • Epistemic status: this is a design-confirmed pattern, not a stress-tested subsystem

Epistemic status: inconclusive. The test validated that the pattern works in favorable conditions, not that it scales to complex graphs or fails gracefully under adversity.

Evidence & receipt
  • file2.1.16/tests/combo-04-skill-templates/TEST-RESULTS.md
◇ ed25519 receipt
idprimitive_6bb31cc51e6902c5c9fda99e
alged25519
pubkey9b87705613b1e2fd064d57fa75a6b679d2856ceafad6b1daa8f982493871b6dd
sig9353c20138ca3dbc344bb5a9011c916531957cbb0a8d98093576ee48686c56c7a4caada5324b3178639920015b861419fc70782bd78059113503705bf9639f09

Signed with an ed25519 key held off the repo. Anyone can verify against the published public key; nobody without the secret key can forge it. Click verify: it recomputes the signature in your browser. The signature proves integrity and authorship of this exact content — not a third-party timestamp or that the underlying claim is objectively true. signedAt is when the @f3/attest pipeline ran, not when the work happened; the evidence refs carry the source dates.

Connected