Preamble

Environment Detection

mkdir -p ~/.omni-skills/sessions
_PROACTIVE=$(~/.claude/skills/superomni/bin/config get proactive 2>/dev/null || echo "true")
_BRANCH=$(git branch --show-current 2>/dev/null || echo "unknown")
_TEL_START=$(date +%s)
echo "Branch: $_BRANCH | PROACTIVE: $_PROACTIVE"

PROACTIVE Mode

If PROACTIVE is false: do NOT proactively suggest skills. Only run skills the user explicitly invokes. If you would have auto-invoked, say: "I think [skill-name] might help here — want me to run it?" and wait.

Completion Status Protocol

Report status using one of these at the end of every skill session:

• DONE — All steps completed. Evidence provided.
• DONE_WITH_CONCERNS — Completed with issues. List each concern explicitly.
• BLOCKED — Cannot proceed. State what blocks you and what was tried.
• NEEDS_CONTEXT — Missing information. State exactly what you need.

Auto-Advance Rule

Pipeline stage order: THINK → PLAN → REVIEW → BUILD → VERIFY → SHIP → REFLECT

REVIEW is the only human gate. All other stages auto-advance on DONE.

Status	At REVIEW stage	At all other stages
DONE	STOP — present review summary, wait for user input (Y / N / revision notes)	Auto-advance — print [STAGE] DONE → advancing to [NEXT-STAGE] and immediately invoke next skill
DONE_WITH_CONCERNS	STOP — present concerns, wait for user decision	STOP — present concerns, wait for user decision
BLOCKED / NEEDS_CONTEXT	STOP — present blocker, wait for user	STOP — present blocker, wait for user

When auto-advancing:

1. Write the session artifact to docs/superomni/
2. Print: [STAGE] DONE → advancing to [NEXT-STAGE] ([skill-name])
3. Immediately invoke the next pipeline skill

Session Continuity

When the user sends a follow-up message after a completed session, before doing anything else:

1. Scan for prior session context:

   ls docs/superomni/specs/spec-*.md docs/superomni/plans/plan-*.md docs/superomni/ .superomni/ 2>/dev/null | head -20
   git log --oneline -3 2>/dev/null
   

   To find the latest spec or plan:

   _LATEST_SPEC=$(ls docs/superomni/specs/spec-*.md 2>/dev/null | sort | tail -1)
   _LATEST_PLAN=$(ls docs/superomni/plans/plan-*.md 2>/dev/null | sort | tail -1)
   

2. If context exists → re-engage the skill framework. Pick the skill that matches the current stage (see workflow skill for stage → skill mapping) and announce: "Continuing in superomni mode — picking up at [stage] using [skill-name]."
3. If no context → treat as a fresh session and offer the relevant skill from the Quick Reference table in using-skills/SKILL.md.

Question Confirmation Protocol

When asking the user a question, match the confirmation requirement to the complexity of the response:

Question type	Confirmation rule
Single-choice — user picks one option (A/B/C, 1/2/3, Yes/No)	The user's selection IS the confirmation. Do NOT ask "Are you sure?" or require a second submission.
Free-text input — user types a value and presses Enter	The submitted text IS the confirmation. No secondary prompt needed.
Multi-choice — user selects multiple items from a list	After the user lists their selections, ask once: "Confirm these selections? (Y to proceed)" before acting.
Complex / open-ended discussion — back-and-forth clarification	Collect all input, then present a summary and ask: "Ready to proceed with the above? (Y/N)" before acting.

Rule: never add a redundant confirmation layer on top of a single-choice or text-input answer.

Custom Input Option Rule: Whenever you present a predefined list of choices (A/B/C, numbered options, etc.), always append a final "Other" option that lets the user describe their own idea:

  [last letter/number + 1]) Other — describe your own idea: ___________

When the user selects "Other" and provides their custom text, treat that text as the chosen option and proceed exactly as you would for any other selection. If the custom text is ambiguous, ask one clarifying question before proceeding.

Context Window Management

Load context progressively — only what is needed for the current phase:

Phase	Load these	Defer these
Planning	Latest docs/superomni/specs/spec-*.md, constraints, prior decisions	Full codebase, test files
Implementation	Latest docs/superomni/plans/plan-*.md, relevant source files	Unrelated modules, docs
Review/Debug	diff, failing test output, minimal repro	Full history, specs

If context pressure is high: summarize prior phases into 3-5 bullet points, then discard raw content.

Output Directory

All skill artifacts are written to docs/superomni/ (relative to project root). See the Document Output Convention in CLAUDE.md for the full directory map.

Feedback Signal Protocol

Agent failures are harness signals — not reasons to retry the same approach:

• 1 failure → retry with a different approach
• 2 failures → surface to user: "Tried [A] and [B], both failed. Recommend [C]."
• 3 consecutive failures → STOP. Report BLOCKED. Treat as a harness deficiency signal. Recommended: invoke harness-engineering skill to update the harness before retrying.
• Uncertain about security → STOP and report NEEDS_CONTEXT
• Scope exceeds verification capacity → STOP and flag blast radius

It is always OK to stop and say "this is too hard for me." Escalation is expected, not penalized.

Performance Checkpoint

After completing any skill session, run a 3-question self-check before writing the final status:

1. Process — Did I follow all defined phases? If any were skipped, state why.
2. Evidence — Is every claim backed by a test result, command output, or file reference? If not, gather the missing evidence now.
3. Scope — Did I stay within the task boundary? If I touched files outside the original scope, flag them explicitly.

If any answer is NO, address it before reporting DONE. If it cannot be addressed, report DONE_WITH_CONCERNS and name the gap.

For a full performance evaluation spanning the entire sprint, use the self-improvement skill.

Telemetry (Local Only)

_TEL_END=$(date +%s)
_TEL_DUR=$(( _TEL_END - _TEL_START ))
~/.claude/skills/superomni/bin/analytics-log "SKILL_NAME" "$_TEL_DUR" "OUTCOME" 2>/dev/null || true

Nothing is sent to external servers. Data is stored only in ~/.omni-skills/analytics/.

Executing Plans

Goal: Execute a written implementation plan precisely, with verification at each stage — running independent steps in parallel to minimize elapsed time.

Iron Laws

1. Dependencies First, Then Parallelize Never execute a step before its dependencies are complete. But DO run all independent steps in parallel within a wave — never serialize work that can be parallelized.

2. Evaluate Before Advancing Every wave must pass an evaluation gate before the next wave begins. A wave is not "done" until its outputs are verified — not just executed.

3. Failures Are Harness Signals When a step fails on 3 consecutive attempts using different approaches, stop executing and treat the failure as a harness signal: update the plan, skill, or constraint — then retry. Never brute-force through 3 failed approaches.

Phase 1: Load the Plan

# Find the plan document
ls docs/superomni/plans/plan-*.md 2>/dev/null | sort | tail -1

Read the plan. Confirm:

• Plan document exists and is readable
• Prerequisites are met
• You understand what "done" looks like for each step

Phase 2: Dependency Analysis — Build the Execution Wave Plan

Before executing any step, analyze ALL steps for dependencies:

DEPENDENCY ANALYSIS
─────────────────────────────────
Step 1: [name] — depends on: none
Step 2: [name] — depends on: none
Step 3: [name] — depends on: Step 1
Step 4: [name] — depends on: none
Step 5: [name] — depends on: Step 2, Step 3
Step 6: [name] — depends on: none
...

WAVE EXECUTION PLAN
Wave 1 (parallel): Steps 1, 2, 4, 6  ← 4 agents dispatched simultaneously
Wave 2 (parallel): Steps 3, 5        ← unblocked after Wave 1 completes
Wave 3 (if needed): ...

Est. time: [N waves] instead of [N sequential steps]

Rule: A step is independent if its outputs are not required by any step in the same wave. Rule: Aim for 5–10 steps per wave when sufficient independent steps exist — never artificially group dependent steps to meet this target.

Phase 3: Execute Wave by Wave

For each wave, dispatch all steps in the wave simultaneously, then wait for all to complete before starting the next wave.

Step Execution Protocol

For each step in a wave:

EXECUTING WAVE [N] — [M] STEPS IN PARALLEL
─────────────────────────────────
Steps: [list of step names/numbers]

For each individual step:

Step [N] — [Step Name]
─────────────────────────────────
What: [Description from plan]
Files: [Files to touch]
Involves code changes? [YES / NO]

1. Read — understand what the step requires
2. TDD Check — if this step involves writing or modifying source code: apply the test-driven-development skill (Red → Green → Refactor) before committing any code
3. Do — make the minimum change needed for this step only
4. Verify — run the step's verification criterion
5. Report — confirm step complete or blocked

TDD Integration for Code Steps

Every step that creates or modifies source code must follow this flow:

Step involves code? ─── NO ──→ Execute directly
        │
        YES
        ↓
Write failing test (RED) → confirm it fails
        ↓
Write minimum implementation (GREEN) → confirm test passes
        ↓
Refactor if needed → confirm tests still pass
        ↓
Continue to step verification

If no test framework exists for this project: document what the tests would look like and why they cannot be automated. This is a DONE_WITH_CONCERNS, not a skip.

Step Completion Format

✓ Step N COMPLETE
  Changed: [files modified]
  Evidence: [test output or verification proof]

Step Blocked Format

✗ Step N BLOCKED
  Blocker: [what prevents completion]
  Tried: [what was attempted]
  Options:
    A) [approach 1]
    B) [approach 2]
    C) Skip this step (explain consequences)
    D) Other — describe your own approach: ___________

Phase 4: Wave Evaluation Gate

Before advancing to the next wave, run the evaluation gate:

WAVE [N] EVALUATION GATE
─────────────────────────────────
Steps completed: [list]
Tests passing:   [run: npm test or equivalent]
Regressions:     [any pre-existing tests broken?]
Output contract: [do outputs match what dependent steps expect?]
Gate result:     PASS → proceed to Wave N+1 | FAIL → address before advancing

If the gate FAILS:

1. Identify which step produced the failing output
2. Determine if this is a harness signal (update plan/skill) or an implementation error (fix and re-run)
3. Do NOT advance to the next wave until the gate passes

Spawn the evaluator agent for complex waves or when the gate result is ambiguous.

Phase 5: Mid-Plan Check-ins

After every wave completes, or when scope is expanding:

1. Report progress: "Completed N/M steps (Wave X of Y done)"
2. Flag if actual work diverges from plan
3. Surface any blast radius discovered mid-execution
4. Ask before proceeding if scope has changed

Phase 6: Handling Plan Deviations

If you discover the plan is wrong or incomplete:

1. Stop — do not improvise silently
2. Assess — is this a small mechanical fix or a fundamental issue?
3. Small fix (mechanical, <5 min): note it, fix it, continue
4. Large issue (taste or architectural): surface to user, wait for input

PLAN DEVIATION DETECTED
Step N: [Original plan says X, but actually Y]
Impact: [Low/Medium/High]
Recommendation: [Proposed resolution]
Awaiting: [Your decision before continuing]

Phase 7: Completion

When all steps are done:

PLAN EXECUTION COMPLETE
════════════════════════════════════════
Steps completed:    N/N
Waves executed:     W
Deviations noted:   N
Files changed:      [list]
Tests passing:      [output]
Status:             DONE | DONE_WITH_CONCERNS
Concerns (if any):
  - [concern 1]
════════════════════════════════════════

Save Execution Results Document

After completing execution, save the results as a Markdown document:

_EXEC_DATE=$(date +%Y%m%d-%H%M%S)
_EXEC_BRANCH=$(git branch --show-current 2>/dev/null | tr '/' '-' || echo "unknown")
_EXEC_FILE="execution-${_EXEC_BRANCH}-${_EXEC_DATE}.md"
mkdir -p docs/superomni/executions
cat > "docs/superomni/executions/${_EXEC_FILE}" << EOF
# Execution Results: ${_EXEC_BRANCH}

**Date:** ${_EXEC_DATE}
**Branch:** ${_EXEC_BRANCH}

[Paste the full PLAN EXECUTION COMPLETE block here]

## Wave Log

[Paste wave-by-wave summary: steps in each wave, outcomes]

## Steps Log

[Paste all step completion/blocked entries here]
EOF
echo "Execution results saved to docs/superomni/executions/${_EXEC_FILE}"

Write the full execution log (wave plan, all step outcomes + the final PLAN EXECUTION COMPLETE block, formatted as Markdown) to docs/superomni/executions/execution-[branch]-[session]-[date].md. This file serves as the permanent record of the execution run for the user to revisit.

Then trigger the verification skill.