feat: artifact pipeline — typed artifact flow between workflow stages (v0.7.0)

[nutt-adam]

Summary

• Artifact Pipeline: Prompt steps can now capture and pass artifacts between workflow stages via artifact_glob and artifact_name fields
• inject_files template expansion: {{output.step_id.path}} references allow artifacts from prior steps to be injected into downstream agents
• Dashboard artifact labels: Flow connectors show artifact names during active runs
• Dry-run validation: Validates gstack-slug availability when {slug} is used in artifact_glob
• Variable interpolation: {slug}, {workspace}, {agent}, ~ expanded in glob patterns

Test Coverage

All new code paths have test coverage. 15 new tests added (326 → 341).

CODE PATH COVERAGE
===========================
[+] src/config/mod.rs — validation
    ├── [★★★ TESTED] artifact_glob + artifact_name parse
    ├── [★★★ TESTED] glob without name rejects
    ├── [★★★ TESTED] name without glob rejects  
    ├── [★★★ TESTED] requires step id
    ├── [★★★ TESTED] requires wait_for_idle
    └── [★★★ TESTED] invalid chars in name rejects

[+] src/automation/mod.rs — capture pipeline
    ├── [★★★ TESTED] expand_artifact_glob ({workspace}, {agent})
    ├── [★★★ TESTED] tilde expansion
    ├── [★★★ TESTED] snapshot_artifact_glob
    ├── [★★★ TESTED] capture picks newest new file
    ├── [★★★ TESTED] zero new files fails
    ├── [★★★ TESTED] multiple files picks newest
    ├── [★★★ TESTED] store_artifact_output
    ├── [★★★ TESTED] render_template with artifact path
    └── [★★★ TESTED] gstack-slug missing → actionable error

COVERAGE: 15/15 paths tested (100%)

Pre-Landing Review

8 issues found and auto-fixed during /review:

• [AUTO-FIXED] inject_files template expansion produced broken double-paths → rewrote with raw string + absolute path handling
• [AUTO-FIXED] artifact_name not validated against [A-Za-z0-9_-] (Codex P3) → added character validation
• [AUTO-FIXED] artifact steps without wait_for_idle could race (Codex P1) → added validation
• [AUTO-FIXED] store_artifact_output returned .raw path, breaking resume (Codex P2) → returns canonical JSON path
• [AUTO-FIXED] snapshot_artifact_glob swallowed errors (Claude [v0.2] Policy enforcement parity across runtimes #6) → returns Result
• [AUTO-FIXED] resolve_gstack_slug ignored exit code (Claude [v0.2] Cost and budget guardrails #8) → checks status.success()
• [AUTO-FIXED] file_name().unwrap_or_default() could produce empty filename (Claude [v0.2] Operator install and validation path #9) → returns error
• [AUTO-FIXED] unsafe env mutation in test without #[serial] (Claude [v0.2] Event stream (SSE/WebSocket) #5) → added #[serial_test::serial]

Adversarial Review

Large-tier (840 lines): Claude structured + Codex structured + Claude adversarial + Codex adversarial.
All high-confidence findings addressed (Codex P1-P3 + Claude #2,5,6,8,9).

Test plan

• All Rust tests pass (341 unit + 1 integration)
• cargo clippy passes with no warnings
• cargo fmt --check passes
• Dry-run validates artifact pipeline workflow correctly

🤖 Generated with Claude Code

Summary by CodeRabbit

• New Features

  • Steps can now capture generated artifacts (files) and pass them to downstream steps using glob patterns and template references.
  • Dashboard visualization now displays artifact names flowing between workflow stages during active runs.
  • Artifact glob patterns support variable interpolation: {slug}, {workspace}, {agent}, and ~ (home directory).
  • Dry-run mode validates artifact configurations before execution.

• Documentation

  • Updated changelog and README with artifact capture capabilities and configuration details.

[coderabbitai]

Warning

Rate limit exceeded

@nutt-adam has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 4 minutes and 10 seconds before requesting another review.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: d4dbc867-cbd7-4e2e-bd42-98861949ef1e

📥 Commits

Reviewing files that changed from the base of the PR and between 4ecd945 and 189d81f.

📒 Files selected for processing (5)

• CHANGELOG.md
• README.md
• dashboard/app.js
• dashboard/style.css
• src/automation/mod.rs

📝 Walkthrough

Walkthrough

This PR introduces artifact-carrying behavior across workflow stages. Steps can now capture files matching a glob pattern (with variable expansion) and register them as outputs, which downstream steps can reference in file injection and templates. Changes include configuration schema additions, executor-level snapshotting and capture logic, dry-run validation, and dashboard visualization of artifacts between stages.

Changes

Cohort / File(s)	Summary
Metadata & Documentation CHANGELOG.md, Cargo.toml, README.md	Version bump to 0.7.0, added glob dependency, documented new artifact capture fields and template references.
Configuration Schema src/config/mod.rs	Added artifact_glob and artifact_name optional fields to WorkflowStepConfig::Prompt with validation rules: fields must be paired, require step id and wait_for_idle: true, and artifact_name must be ASCII alphanumeric plus underscore/hyphen.
Core Automation Logic src/automation/mod.rs	Implemented artifact snapshotting before step execution and capture after completion, supporting variable expansion (~, {workspace}, {agent}, {slug}). Enhanced inject_files template rendering and added validate_gstack_slug_available() for slug validation. Extended ResolvedStep::Prompt with inject_files_raw, artifact_glob, artifact_name.
CLI & Dry-Run src/cli/run.rs	Added slug validation during dry-run for steps with {slug} in artifact glob. Extended DryRunStep::Prompt with optional artifact fields and enhanced step summary display.
Dashboard Visualization dashboard/app.js, dashboard/style.css	Added artifactName tracking on step start events, introduced artifactBetweenStages() helper to detect artifacts flowing between stages, and conditionally render artifact labels on pipeline connectors with styling for absolute positioning and monospaced appearance.

Sequence Diagram

sequenceDiagram
    participant Executor
    participant FileSystem as File System
    participant OutputRegistry as Output Registry
    participant Dashboard

    Executor->>FileSystem: Snapshot files matching artifact_glob (pre-step)
    Note over Executor,FileSystem: Expand variables: ~, {workspace}, {agent}, {slug}
    
    Executor->>Executor: Execute prompt step
    
    Executor->>FileSystem: Scan for new files matching expanded glob (post-step)
    FileSystem-->>Executor: Return newest matching file
    
    Executor->>OutputRegistry: Register artifact as step output<br/>(JSON + .raw copy)
    OutputRegistry-->>Executor: Output stored
    
    Executor->>Dashboard: Emit step.started event with artifact_name
    Dashboard->>Dashboard: Query artifactBetweenStages(prevStage, stage)
    Dashboard->>Dashboard: Render artifact label on pipeline connector

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~45 minutes

Possibly related PRs

• [auto] #84 feat: web dashboard with live agent terminal views #90: Modifies dashboard pipeline rendering and SSE event handling — the artifact-label additions in this PR build directly on the dashboard machinery introduced there.
• [auto] #67 phase 1: stabilize fresh prompt-step startup on brand-new choir runs #71: Extends ResolvedStep::Prompt shape by adding new step-level fields (this PR adds artifact fields; retrieved PR adds startup grace period) — both touch the same enum variant and workflow step resolution.

Poem

🐰 Artifacts now hop between stages,
Glob patterns capture files through ages,
Variables expand with a twitch,
Dashboard labels make flows rich,
Tutti workflows grow their wages! ✨

🚥 Pre-merge checks | ✅ 1 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 78.95% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check	❓ Inconclusive	The description is comprehensive and addresses most template requirements, but the Versioning section checklist items are not explicitly checked off and SemVer choice selection is not filled in.	Explicitly check off the Versioning checklist items and fill in the SemVer choice (MINOR is appropriate for new capability) and version number field.

✅ Passed checks (1 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically describes the main change: introducing an artifact pipeline feature for typed artifact flow between workflow stages, with version bump to 0.7.0.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/artifact-pipeline

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 7

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@CHANGELOG.md`:
- Around line 6-12: The changelog uses inconsistent terminology and a mismatched
template key; update the Artifact Pipeline entry to consistently refer to
"skill" (not "prompt steps") and change the template example to the README's
placeholder style by using the `artifact_name` key (and keep `artifact_glob` and
`artifact_name` symbol names intact); also update the inject_files line to show
the supported template reference using `{{output.artifact_name.path}}` and
ensure both descriptions reference "skill" and the `artifact_name` placeholder
consistently so operators know which key to use.

In `@dashboard/app.js`:
- Around line 130-148: artifactBetweenStages currently inspects every run in
appState.runs and can return artifact labels from terminal runs; update the
function to skip any run whose status is not "running" by checking run.status
=== "running" (or !== "running" to continue) before iterating run.steps in
artifactBetweenStages so only active runs contribute labels.

In `@dashboard/style.css`:
- Around line 141-153: The .artifact-label rule uses an incorrect CSS variable
name; replace var(--mono) with the defined variable var(--font-mono) in the
.artifact-label font-family declaration so the label uses the JetBrains Mono
font; update the font-family in the .artifact-label selector accordingly.

In `@README.md`:
- Line 267: The README documents the artifact output interpolation as
{{output.artifact_name.path}} but the changelog uses {{output.step_id.path}},
which will confuse users; update the README text to use the canonical key format
used by the codebase (choose either {{output.artifact_name.path}} or
{{output.step_id.path}}) or explicitly state both supported forms and when to
use each, and ensure the example inject_files snippet and the surrounding
sentence reference the exact template key you choose so docs and changelog
match.

In `@src/automation/mod.rs`:
- Around line 2548-2584: store_artifact_output currently saves a JSON wrapper to
canonical_path and copies the raw artifact to raw_path but returns
StepOutputValue.path = canonical_path, so render_template/inject_files will pick
the JSON wrapper instead of the raw artifact; change the returned/exposed path
to the raw artifact by setting StepOutputValue.path to raw_path (while still
calling save_workflow_output and keeping canonical_path for the persisted JSON),
e.g. keep json_value and canonical_path as-is, perform the std::fs::copy to
produce raw_path, and then return StepOutputValue { path: raw_path, json:
json_value } so {{output.<name>.path}} refers to the raw file that downstream
inject_files should copy.
- Around line 1486-1546: The artifact capture/store logic (using
capture_artifact and store_artifact_output and updating outputs, output_files,
step_results, failed_steps, and success) must be extracted into a small helper
(e.g., capture_and_store_artifact) that accepts the run context (run_id,
step_index, started, artifact_pre_snapshot, artifact_name, project_root, etc.)
and returns a Result<saved_artifact_info, failure_marker>; implement the helper
to perform the sleep, call capture_artifact(expanded_pattern, pre_snap), call
store_artifact_output(self.project_root, &run_id, art_name, &artifact_path),
update outputs/output_files on success or push to failed_steps/step_results and
set success=false on error (constructing the same StepResult shape used
elsewhere); then call this helper from every early-success exit path (the
success continue at the prompt fallthrough and the implement_code
auto-finalization success branches) before returning/continuing so every
successful prompt path registers artifacts.
- Around line 6078-6097: The test
gstack_slug_missing_binary_returns_actionable_error mutates the process HOME and
introduces race conditions; change resolve_gstack_slug to accept an injectable
home directory (e.g., fn resolve_gstack_slug(home: impl AsRef<Path>) ->
Result<...>) or an Option<&Path> defaulting to env::var("HOME") when None,
update all call sites to pass either the real HOME or a test-supplied
non-existent path, and modify the test to call
resolve_gstack_slug("/nonexistent-path-for-test") instead of setting the HOME
env var; ensure signature and uses are updated consistently across the module
and propagate the new parameter through helper functions that locate the
gstack-slug binary.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: aadb3d98-c5aa-4ab3-bf45-4b85da15cbec

📥 Commits

Reviewing files that changed from the base of the PR and between 150da11 and 4ecd945.

⛔ Files ignored due to path filters (1)

• Cargo.lock is excluded by !**/*.lock

📒 Files selected for processing (8)

• CHANGELOG.md
• Cargo.toml
• README.md
• dashboard/app.js
• dashboard/style.css
• src/automation/mod.rs
• src/cli/run.rs
• src/config/mod.rs