feat(cli): surface the harness golden path (af wait, catalog, golden_path help, JSON envelopes)

[AbirAbbas]

Problem

The "golden path" an AI harness follows to drive AgentField agent nodes —
discover → schema → call async → wait → result — exists as primitives but was
never surfaced as one coherent flow:

• af agent help (the machine-friendly interface) taught discovery and
  introspection but never execution: its quick_start omitted af call and
  af tail entirely.
• af call --async printed a bare run-id string even under -o json, so a
  parser asking for JSON got a non-JSON token.
• af call / af ls / af tail against a down control plane emitted a raw Go
  dial error with no guidance, while af agent commands already appended a
  reachability hint.
• af skill install with no argument installed only the first catalog entry —
  the agentfield build skill — never agentfield-use, the drive skill
  that documents the golden loop.
• There was no CLI way to browse installable nodes; the only curated catalog
  lived in the desktop app.

What changed

• golden_path in af agent help — an ordered array of {step, command, purpose} covering doctor → catalog → install → secrets → run → ls/discover →
  call --schema → call --async → wait/tail. af call and af tail
  added to quick_start.
• af wait <run_id> [--timeout <sec, default 600>] — polls the run overview
  (the same /api/v1/agentic/run/:run_id API af agent run --id uses) until the
  run is terminal, prints final status + result as JSON, and maps exit codes:
  0 succeeded, 1 failed/cancelled, 2 timeout.
• af call --async JSON envelope — under an explicit -o json/-o yaml it
  emits {"run_id": "...", "status": "accepted"}. The default/pretty path keeps
  the bare run-id line scripts capture via RUN_ID=$(af call ... --async).
• Shared control-plane-unreachable hint — a transport failure through the
  shared makeRequest now surfaces Control plane not reachable at <url>. Start it with af server or launch the AgentField desktop app. for call/ls/tail/wait.
  A cancelled context is passed through unwrapped.
• af catalog — a new command backed by an in-binary catalog seeded from the
  desktop app's curated list (name, description, install source, docs URL).
  Supports -o json/-o yaml; the human table ends with the hint
  af install <source>.
• af skill install with no name installs the whole catalog — both the build
  and drive skills, via a new skillkit.InstallAll. Explicit af skill install <name> is unchanged; the interactive picker copy now names both skills.

Validation Contract

Behavior (intent)	Test
af agent help JSON contains golden_path with the ordered driving loop; quick_start includes af call and af tail	TestAgentHelpDataShape (agent_helpers_test.go)
af wait on a succeeded run exits 0 and prints JSON with status and result; --timeout produces exit 2	TestRunWait (wait_test.go) — subtests "succeeded run exits 0…" and "times out with exit code 2…" (plus failed→exit 1 and mid-run polling)
af call x.y --async -o json output parses as JSON with run_id	TestRunCallSyncSchemaAndAsync (trigger_commands_test.go)
Unreachable control plane on call/ls/tail/wait mentions af server	TestUnreachableControlPlaneHintAcrossCommands + TestControlPlaneUnreachableErrorMentionsServer (connection_hint_test.go)
af catalog -o json yields ≥5 entries each with name/description/source	TestRunCatalogJSON (catalog_test.go)
Bare af skill install (non-interactive) installs both skills	TestInstallAllInstallsEveryCatalogSkill (skillkit/install_all_test.go)

Test run

Ran the CI gates locally (from .github/workflows/control-plane.yml):

• control-plane/web/client: npm ci && npm run build — pass
• GOFLAGS=-buildvcs=false go build ./... — pass
• go vet ./... — pass
• gofmt -l on changed files — clean
• golangci-lint run on changed packages — no new findings on changed files
• go test ./internal/cli/... ./internal/skillkit/... — pass

Note: one unrelated test, internal/server/TestStartAndStopCoverAdditionalBranches,
fails in this local WSL environment (it binds an admin gRPC server and dials a
stub health endpoint that hangs locally). It reproduces identically on a pristine
origin/main checkout and has no dependency on the packages changed here, so it
is a pre-existing environment artifact, not a regression from this PR.

🤖 Generated with Claude Code

[github-actions]

📊 Coverage gate

Thresholds from .coverage-gate.toml: per-surface ≥ 84%, aggregate ≥ 85%, max per-surface regression ≤ 1.0 pp, max aggregate regression ≤ 0.50 pp.

Surface	Current	Baseline	Δ
control-plane	86.90%	87.40%	↓ -0.50 pp	🟡
sdk-go	92.50%	92.00%	↑ +0.50 pp	🟢
sdk-python	93.82%	93.73%	↑ +0.09 pp	🟢
sdk-typescript	91.05%	90.42%	↑ +0.63 pp	🟢
web-ui	84.75%	84.79%	↓ -0.04 pp	🟡
aggregate	85.54%	85.75%	↓ -0.21 pp	🟡

✅ Gate passed

No surface regressed past the allowed threshold and the aggregate stayed above the floor.

[github-actions]

📐 Patch coverage gate

Threshold: 80% on lines this PR touches vs origin/main (from .coverage-gate.toml:thresholds.min_patch).

Surface	Touched lines	Patch coverage	Status
control-plane	227	94.00%	✅
sdk-go	0	—	➖ no changes
sdk-python	0	—	➖ no changes
sdk-typescript	0	—	➖ no changes
web-ui	0	—	➖ no changes

✅ Patch gate passed

Every surface whose lines were touched by this PR has patch coverage at or above the threshold.