Skip to content

Add tvc-app-builder skill with eval CI#2

Open
natefikru wants to merge 8 commits intomainfrom
feat/tvc-app-builder-skill
Open

Add tvc-app-builder skill with eval CI#2
natefikru wants to merge 8 commits intomainfrom
feat/tvc-app-builder-skill

Conversation

@natefikru
Copy link
Copy Markdown

@natefikru natefikru commented Mar 26, 2026

Summary

  • Adds a Claude Code agent skill (skills/tvc-app-builder/) that guides autonomous agents through the full TVC app lifecycle: project setup, Rust endpoint implementation, testing, OCI container builds, and deployment via the tvc CLI
  • Includes skill eval CI that validates structure and runs trigger accuracy tests on every PR touching skills/
  • CLI documentation aligned with the latest rust-sdk/tvc source code

What the skill teaches an agent

  • How to read and extend the tvc-template project structure
  • Axum route handlers, error handling, shared state patterns
  • Unit tests (Tower oneshot) and e2e tests (Builder harness)
  • OCI container builds with StageX reproducible pipeline
  • Fully autonomous TVC deployment using --json, --no-input, --yes CLI flags
  • Troubleshooting: private container images, wrong app URL domains, binary vs container digest confusion
  • Stateless enclave design (multi-instance load balancing means in-memory state is per-instance)
  • Calling the Turnkey API from TVC apps and deployment scripts

Skill structure

skills/tvc-app-builder/
  SKILL.md                              # Core instructions (v3.2.0)
  scripts/
    eval-triggers.ts                    # Trigger accuracy eval runner
    validate.ts                         # Structural validator
  references/
    tvc-cli-guide.md                    # Full CLI reference with automation flags
    deployment-troubleshooting.md       # Common deployment issues and fixes
    template-architecture.md            # Project layout, middleware, lint rules
    testing-guide.md                    # Unit, e2e, and manual test patterns
    app-examples.md                     # 5 example app designs
    turnkey-api-integration.md          # Turnkey API patterns from TVC apps/scripts
  evals/
    triggers.json                       # 22 positive, 8 negative trigger queries
    evals.json                          # 6 functional eval cases

Test plan

  • npm run validate passes
  • Verify skill-eval.yml CI workflow runs on this PR
  • Test skill by loading it in Claude Code and asking "Build a TVC app"

Adds a Claude Code skill that guides autonomous agents through building,
testing, and deploying TVC enclave applications. Includes a working
timestamp notary service as a demo app.

Skill (skills/tvc-app-builder/):
- SKILL.md with full build-test-deploy workflow using autonomous CLI flags
- TVC CLI reference covering --json, --no-input, --yes for non-interactive use
- Deployment troubleshooting guide (private images, wrong domains, digest confusion)
- Template architecture, testing patterns, and app examples references
- Trigger evals passing at 93.5% accuracy with 0% false positives

Demo app (src/helloworld/):
- POST /notarize: accepts document hash, returns timestamped receipt
- GET /verify/{receipt_id}: looks up receipt by ID, 404 if missing
- GET /stats: total notarized document count
- Shared state via Arc<RwLock<AppState>> with proper error handling
- 14 unit tests + 4 e2e tests, all passing

CI (.github/workflows/skill-eval.yml):
- Validates skill structure on skills/** changes
- Runs trigger evals with 90% threshold
@socket-security
Copy link
Copy Markdown

socket-security Bot commented Mar 26, 2026

Review the following changes in direct dependencies. Learn more about Socket for GitHub.

Diff Package Supply Chain
Security
Vulnerability Quality Maintenance License
Addednpm/​tsx@​4.21.01001008185100
Addednpm/​gray-matter@​4.0.39910010083100
Addednpm/​typescript@​5.9.31001009010090

View full report

@natefikru natefikru marked this pull request as draft March 26, 2026 03:53
Updated skill to match actual TVC CLI behavior observed during a real
deployment. Removed non-existent global flags (--json, --no-input, --yes),
corrected deploy approve to use --dangerous-skip-interactive, replaced
JSON parsing instructions with grep-based output parsing and cached config
reads, added ghcr.io auth and package visibility steps, documented
enclave provisioning delay and dev environment URL pattern.
…global flags

Restored --json, --no-input, --yes global flags and login flags (--alias,
--api-env, --skip-api-key-wait) to match the updated TVC CLI. Kept the
deployment workflow improvements from v3.1.0: ghcr.io authentication step,
package visibility instructions, enclave provisioning delay documentation,
and dev environment URL pattern.
…st-sdk

Remove all src/ changes (router, e2e tests, Cargo files) to keep this
as a skill-only PR. Fix CLI documentation to match actual rust-sdk/tvc
source: remove nonexistent --operator-key-file global flag, remove
nonexistent --skip-api-key-wait login flag, mark tvc app list as
unimplemented.
@natefikru natefikru changed the title Add tvc-app-builder skill v3.0.0 with notary demo and CI Add tvc-app-builder skill with eval CI Mar 26, 2026
@natefikru natefikru marked this pull request as ready for review March 26, 2026 22:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant