pipeline

name	pipeline
description	Invoke and resume YAML-defined pipelines by name — /pipeline auto-dev runs the full release pipeline
scope	harness
user-invocable	true
effort	high
argument-hint	<pipeline-name> | resume | (no args to list available)
source	{"type":"external","origin":"github","url":"https://github.com/baekenough/baekenough-skills","version":"1.0.0"}

/pipeline — Pipeline Invocation

Usage

/pipeline auto-dev          # Run the auto-dev pipeline
/pipeline                   # List available pipelines
/pipeline resume            # Resume a halted pipeline

Behavior

Workflow File Locations

workflows/*.yaml is the Codex pipeline invocation surface used by /pipeline <name> in this repository. templates/workflows/*.yaml is its packaged template mirror and must remain content-identical.

.codex/skills/pipeline/workflows/*.yaml is a skill-local reference copy retained for compatibility with older skill bundles and detailed examples. It is not the primary repo-root invocation surface unless a runtime explicitly resolves workflows relative to this skill directory.

When changing pipeline behavior, update the active repo-root workflow first, update its template mirror in the same change, and then update skill-local reference copies when they describe the same behavior.

List Mode (no arguments or --list flag)

Execute these steps to display available pipelines:

1. Scan built-in pipelines: Use Glob("workflows/*.yaml") (NOT templates/) to find all pipeline definitions
2. Extract metadata: For each YAML file found, use Bash to extract name and description:

   for f in workflows/*.yaml; do
     name=$(grep -m1 '^name:' "$f" | sed 's/^name: *//' | tr -d '"')
     desc=$(grep -m1 '^description:' "$f" | sed 's/^description: *//' | tr -d '"')
     echo "  $name — $desc"
   done
   

3. Scan template pipelines: Use Glob("templates/workflows/*.yaml") for template examples
4. Display formatted output:

   Available pipelines:
     {name} — {description}
     {name} — {description}
   
   Template pipelines (in templates/workflows/):
     {name} — {description}
   

5. If no pipelines found, display: "No pipelines found in workflows/ directory."
6. If YAML parsing fails for a file, skip it and show: {filename} — (parse error, skipped)

Run Mode (with pipeline name)

1. Validate pipeline exists: workflows/{name}.yaml
2. Load and validate YAML structure:
   • Required fields: name, description, steps[]
   • Each step has either skill: or prompt: (not both)
   • Referenced skills exist in .codex/skills/
   • Skill names must match ^[a-z0-9-]+$ (kebab-case only) — reject path traversal attempts
3. Announce: [Pipeline] Starting {name} — {step_count} steps
4. Execute steps top-to-bottom:
   • Skill steps (skill: name): Invoke via Skill tool — Skill(skill: "{name}")
   • Prompt steps (prompt: text): Execute the described action using appropriate agents/tools
   • Foreach steps (foreach: collection): Iterate over collection from previous step output
   • Parallel steps (parallel: [step1, step2]): Execute contained steps concurrently using Agent tool. Each parallel step runs as an independent Agent. Max 4 concurrent per R009. Steps within a parallel block MUST be independent (no shared state, no sequential dependencies). Dependencies between parallel and non-parallel steps use depends_on: field.
   • Claude compatibility Agent calls only (R010 “Delegated Permission Ownership”): Pass mode: "bypassPermissions" when the active Claude session uses bypass permissions. Native Codex spawn_agent has no mode parameter; use the installed agent_type and active Codex runtime permissions instead.
5. Report completion or failure

Resume Mode (/pipeline resume)

1. Scan /tmp/.codex-pipeline-*-{PPID}.json for state files
2. If none found: "No halted pipelines found."
3. If found: display pipeline name, failed step, error message
4. Options:
   • Retry — Re-execute the failed step
   • Skip — Mark failed step as skipped, continue to next
   • Abort — Delete state file, cancel pipeline
5. On resume: execute from the failed step

State Tracking

Track per-step state:

{
  "pipeline": "{name}",
  "started": "ISO-8601",
  "status": "running|completed|halted",
  "current_step": 0,
  "steps": [
    {"name": "triage", "status": "completed", "duration_ms": 5000},
    {"name": "plan", "status": "running"}
  ]
}

State saved to /tmp/.codex-pipeline-{name}-{PPID}.json on failure.

Phase Token Spend Tracking

For release pipelines such as auto-dev, record an advisory token-spend estimate at every phase boundary. This is intentionally lightweight and does not require provider billing APIs.

• State file: /tmp/auto-dev-spend-{PPID}.json
• Estimate: (input_chars + output_chars) / 4, rounded to the nearest integer
• Required fields per phase: name, started_at, completed_at, input_chars, output_chars, estimated_tokens
• Final report: print a phase table and total estimated tokens before the follow-up step

If exact usage events are available from the runtime, prefer them and set token_source: "runtime". Otherwise set token_source: "estimated". Missing spend data must not block a release; it should be reported as an observability gap.

Parallel Execution

Pipeline steps can be grouped for parallel execution:

steps:
  - name: phase-1
    parallel:
      - name: task-a
        skill: skill-a
        description: First independent task
      - name: task-b
        skill: skill-b
        description: Second independent task
  - name: phase-2
    skill: next-step
    depends_on: phase-1

Parallel Rules

• Max 4 concurrent steps per parallel block (R009 hard cap)
• Steps within a parallel block MUST be independent
• depends_on enforces ordering between blocks
• Each parallel step is spawned as a separate Agent tool call in the SAME message
• Preserve the session permission posture only for Claude compatibility Agent calls; native Codex spawn_agent follows the active runtime permissions and never accepts mode
• If any parallel step fails with error: halt-and-report, all remaining steps in the block are cancelled
• State tracking records each parallel step individually

Parallel State Format

{
  "name": "phase-1",
  "type": "parallel",
  "status": "running",
  "children": [
    {"name": "task-a", "status": "completed", "duration_ms": 5000},
    {"name": "task-b", "status": "running"}
  ]
}

Error Handling

• Pipeline not found → list available pipelines with suggestion
• YAML parse error → report with line number
• Step failure (error: halt-and-report) → stop execution, save state, report failure with context
• All file writes delegated to subagents per R010