All skills
simota avatar

/nexus

@c805268
by shingo imotasimota/agent-skills85 stars
15

Orchestrating multi-specialist task chains and scope-adaptive product delivery: classifies intent, selects and executes the minimum viable chain, aggregates results, and verifies acceptance criteria. For multi-domain tasks, build-first delivery, and product lifecycle execution.

Use this Skill: https://skilld.dev/gh/simota/agent-skills/nexus

This session only. Nothing lands on disk.

referenceoutput-formats.md

≈3.9k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Nexus Output Formats Reference

Purpose: Canonical final output and handoff templates. Read when: You need exact NEXUS_COMPLETE, handoff, or report formatting.

Contents

  • Envelope Sizing (read first)
  • NEXUS_COMPLETE (AUTORUN)
  • Work Gate Matrix
  • Compact Form (SIMPLE runs)
  • NEXUS_COMPLETE_FULL (AUTORUN_FULL)
  • NEXUS_HANDOFF_V2 (Standard - Required)
  • NEXUS_HANDOFF (Legacy - Deprecated)
  • NEXUS_HANDOFF (Extended - AUTORUN_FULL)
  • _STEP_COMPLETE Format

Final output formats and handoff protocols.


Envelope Sizing

The templates below are the largest form of each envelope, not a per-run floor (_common/OUTPUT_STYLE.md § Conditional Requirements). Pick the form the run actually filled.

Run shape Form
SIMPLE — one agent, one step, no branches, no guardrail events Compact Form below
MEDIUM — 2–3 steps, sequential NEXUS_COMPLETE, empty ledgers collapsed to one line
COMPLEX — 4+ steps, parallel branches, or any guardrail event NEXUS_COMPLETE_FULL

Collapse rules, binding on every form:

  • A ledger with no entries is one line, not a header plus a column-header-only table: Decisions: none · Residuals: none · Guardrail events: none.
  • Acceptance Provenance stays a table whenever there is ≥1 criterion — this is the one ledger that must not collapse, because a criterion silently absent is the defect it exists to catch. With a single criterion, one row is the whole table.
  • ### Changes lists files, not narration. One line per file; no restatement of what the diff shows.
  • Never emit N/A, "none identified", or placeholder rows to satisfy a template.
  • The completion sweep line is never dropped, in any form, including Compact. It carries two counts: residue found, and files written / files evidenced. While the second pair differs the status is not SUCCESS — a file in neither the evidence nor the Residual Ledger is an unverified change with nothing declaring it so (_common/OPERATIONAL.md § Completion Contract).

NEXUS_COMPLETE (AUTORUN)

## NEXUS_COMPLETE
Task: [Task name]
Type: [BUG|FEATURE|REFACTOR|...]
Chain: [Executed chain]
Fallback: [fallback_taken: compass-invoked | architect-invoked | neither — reason: <reason>] (only when the LADDER step ran; omit line otherwise)

### Changes
- [File1]: [Change description]
- [File2]: [Change description]

### Verification
- Tests: [PASS/FAIL + details]
- Build: [status]

### Acceptance Provenance
| Criterion | Class | Evidence / gap |
|-----------|-------|----------------|
| [AC or derived criterion] | verified \| partial \| missed \| dropped(DEC-n) | [observed evidence, or the precise gap] |
| [prohibited outcome] | held \| violated \| unverified | [evidence it did not occur, or why unobservable] |

### Decision Ledger
- [DEC-n (class)]: [decision] — [why] ([interpretation entries first; omit section if empty])

### How to Verify
1. [Verification step 1]
2. [Verification step 2]

### Risks / Residual Ledger
- [Remaining risks + any UNVERIFIED claims]

| ID | Residual | Class | Blocker / owner | Marker location | Route |
|----|----------|-------|-----------------|-----------------|-------|
| RES-n | [what is not done] | blocked-external \| gate-pending \| out-of-contract \| budget-exhausted \| user-declined | [named blocker] | [file:line `#TODO(agent):`, or `none`] | [recipe/agent that finishes it] |

Completion sweep: [command run] — [N hits: each mapped to a RES-n, or `pre-existing`]; [X] changed / [Y] evidenced (`scanned, 0 hits; 7 changed / 7 evidenced` when clean)

Acceptance Provenance covers every intent-contract criterion (none silent) and every prohibited outcome on its own axis — a prohibition is held only with evidence that the forbidden result did not occur; unverified is honest, held by assumption is not, and violated caps the run at FAILED. Decision Ledger and evidence rules per reference/autonomy-quality-protocol.md (Q2, Q4–Q6, Q10, Q15).

The Residual Ledger replaces free-text "Recommended follow-ups": every leftover carries a Q17 class, and an untyped residual is a defect that caps status at PARTIAL (Q16–Q19). Rows and in-artifact #TODO(agent): markers bind bidirectionally — no orphan markers, no orphan rows. Omit the table only when the sweep is clean, and keep the sweep line either way.


Work Gate Matrix

Every spoke emits WORK_GATE (_common/WORK_GATE.md). Nexus renders them as a matrix — skills down, axes across — and never as a rollup. No row total, no column average, no "overall ★". The reader's question is which skill, which axis, and any aggregate erases exactly that.

### Work Gate

| Skill   | IN    | FIT   | EVD   | OUT   | RSK  | CLR   |
|---------|-------|-------|-------|-------|------|-------|
| scout   | ★★☆☆☆ | ★★★★★ | ★★★★☆ | ★★★☆☆ | pass | ★★★★★ |
| builder | ★★★★☆ | ★★★★★ | ★★★☆☆ | ★★★☆☆ | pass | ★★★★☆ |
| radar   | ★★★★☆ | ★★★★★ | ★★★★★ | ★★★★☆ | pass | ★★★★☆ |

- scout / IN ★★ — repro steps absent from the report; reproduced from the stack trace alone
- builder / EVD ★★★ — the latency claim is labelled UNVERIFIED

Rules:

  • Every cell at ★★★☆☆ or below gets one line under the table naming the reason. A low cell with no line is an incomplete report.
  • n/a cells render n/a, never blank, never as ★1, never merged into ★5.
  • Any RSK: risk blocks NEXUS_COMPLETE and is reported at the top of the envelope, not as a cell the reader has to find. RSK is never starred.
  • A spoke that emitted no gate renders as a — row with "no gate emitted". Silence is a finding about the spoke, not an empty cell.
  • Skip-tier spokes contribute only the axes they emitted; the rest render · (not asserted at this tier) rather than stars.
  • IN is read as a column, not per row. A chain where IN degrades from ★★★★☆ at the head to ★★☆☆☆ three spokes down is a handoff defect, and it is invisible in any single spoke's gate.
  • Rework is never invisible. A spoke that was sent back under _common/HANDOFF.md § Handoff Admission Gate is marked ↺ beside its name, with the refusing skill and the named missing item on a line below the table. Nexus is the arbiter of the one-bounce-per-edge rule; a chain that silently reran a step reports a smaller cost than it paid.
  • A second failure on the same condition is a blocker, not a third attempt — it surfaces at the top of the envelope as a typed residual plus the question for the user.

Compact Form (SIMPLE runs)

## NEXUS_COMPLETE
Task: [Task name] · Type: [TYPE] · Chain: [Agent]

### Changes
- [file]: [change]

### Verification
- [Tests / build / check]: [result, or `UNVERIFIED — <why>`]

### Acceptance
| Criterion | Class | Evidence / gap |
|-----------|-------|----------------|
| [AC] | verified \| partial \| missed | [evidence] |

Decisions: none | Residuals: none
Completion sweep: [command] — scanned, 0 hits; 1 changed / 1 evidenced

Escalate out of Compact the moment any of these is true: a DEC-n was taken, a residual exists, a guardrail fired, a criterion is partial/missed, or a prohibited outcome is unverified. Those are the cases the long form exists for — everything else is scaffolding.


NEXUS_COMPLETE_FULL (AUTORUN_FULL)

## NEXUS_COMPLETE_FULL
Task: [Task name]
Type: [BUG|FEATURE|REFACTOR|...]
Mode: AUTORUN_FULL
Complexity: [SIMPLE|MEDIUM|COMPLEX]

### Execution Summary
- Total Steps: [N]
- Parallel Branches: [N branches if any]
- Duration: [Phases completed]
- Recovery Actions: [N if any]

### Chain Executed
Sequential: [Agent1] → [Agent2] → [Agent3]
Parallel (if any):
  Branch A: [Agent4] → [Agent5]
  Branch B: [Agent6] → [Agent7]
  Merge: [Agent8]

### Changes
- [File1]: [Change description]

### Guardrail Events
| Step | Level | Trigger | Action | Result |
|------|-------|---------|--------|--------|
| 3/7 | L2 | test_failure | auto_fix | SUCCESS |

### Verification
- Tests: [PASS/FAIL + details]
- Build: [status]
- Security: [Sentinel result if applicable]
- Final Guardrail: [L2 CHECKPOINT result]

### Context Summary
- Goal: [Original goal + non-goals from the intent contract]
- Acceptance: [per-criterion — see Acceptance Provenance table; never a blanket "all criteria met"]
- Key Decisions: [Decision Ledger — DEC-n entries, interpretation class first]

### Acceptance Provenance
| Criterion | Class | Evidence / gap |
|-----------|-------|----------------|
| [AC or derived criterion] | verified \| partial \| missed \| dropped(DEC-n) | [observed evidence, or the precise gap] |
| [prohibited outcome] | held \| violated \| unverified | [evidence it did not occur, or why unobservable] |

### How to Verify
1. [Verification step 1]
2. [Verification step 2]

### Risks / Residual Ledger
- [Remaining risks + any UNVERIFIED claims]

| ID | Residual | Class | Blocker / owner | Marker location | Route |
|----|----------|-------|-----------------|-----------------|-------|
| RES-n | [what is not done] | blocked-external \| gate-pending \| out-of-contract \| budget-exhausted \| user-declined | [named blocker] | [file:line `#TODO(agent):`, or `none`] | [recipe/agent that finishes it] |

Completion sweep: [command run] — [N hits: each mapped to a RES-n, or `pre-existing`]; [X] changed / [Y] evidenced (`scanned, 0 hits; 7 changed / 7 evidenced` when clean)

### Rollback (if needed)
- Rollback available: [Yes/No]
- Command: [git checkout / restore command]

NEXUS_HANDOFF_V2 (Standard - Required)

All agents MUST use V2 format with confidence scoring.

Compliance Levels: Level 1 (Minimal) requires only step, agent, status, summary, next_agent, next_action — confidence is inferred from status (see handoff-validation.md Compliance Levels). Level 2 adds confidence as a single number. Level 3 (Full/Claude default) adds confidence_breakdown with 3 axes.

## NEXUS_HANDOFF
step: [X/Y]
agent: [AgentName]
status: [SUCCESS|PARTIAL|BLOCKED|FAILED]

# REQUIRED: Confidence scoring for auto-routing
confidence: 0.XX  # Overall score (0.0-1.0)
confidence_breakdown:
  task_completion: 0.XX   # How complete is the work
  output_quality: 0.XX    # Quality of artifacts produced
  next_step_clarity: 0.XX # How clear is the next step

summary: |
  [1-3 line summary of work completed]

key_findings:
  - [Finding 1]
  - [Finding 2]

artifacts:
  - type: [file|command|link]
    path: [path]
    description: [what it is]

risks:
  - [Risk 1]
  - [Risk 2]

open_questions:
  - blocking: [true|false]
    question: [Question]

pending_confirmations:  # Only if status == BLOCKED
  - trigger: [INTERACTION_TRIGGER]
    question: [Question]
    options: [List]
    recommended: [Option]

user_confirmations:
  - question: [Previous Q]
    answer: [User's A]

next_agent: [AgentName|DONE]
next_action: [CONTINUE|MERGE|VERIFY|ESCALATE|ABORT]
reason: [Why this next step]

Auto-Routing Rules

Confidence Status Action
>= 0.75 SUCCESS Auto-route to next_agent
0.50-0.74 SUCCESS/PARTIAL Route with logged assumptions
< 0.50 any Pause for user input
any BLOCKED Present pending_confirmations
any FAILED Execute recovery chain or escalate

See reference/handoff-validation.md for full validation rules.


NEXUS_HANDOFF (Legacy - Deprecated)

## NEXUS_HANDOFF
- Step: [X/Y]
- Agent: [AgentName]
- Summary: 1-3 lines
- Key findings / decisions:
  - ...
- Artifacts (files/commands/links):
  - ...
- Risks / trade-offs:
  - ...
- Open questions (blocking/non-blocking):
  - ...
- Pending Confirmations:
  - Trigger: [INTERACTION_TRIGGER name]
  - Question: [Question]
  - Options: [Available options]
  - Recommended: [Recommended option]
- User Confirmations:
  - Q: [Previous question] → A: [User's answer]
- Suggested next agent: [AgentName] (reason)
- Next action: CONTINUE

Note: Legacy format will be auto-migrated to V2 with inferred confidence. See handoff-validation.md for migration rules.


NEXUS_HANDOFF (Extended - AUTORUN_FULL)

## NEXUS_HANDOFF
- Step: [X/Y]
- Branch: [branch_id or "main"]
- Agent: [AgentName]
- Summary: 1-3 lines
- Key findings / decisions:
  - ...
- Artifacts (files/commands/links):
  - ...
- Files Modified: [List of files]
- Risks / trade-offs:
  - ...
- Guardrail Events:
  - Level: [L1|L2|L3|L4 or "none"]
  - Trigger: [What triggered if any]
  - Action: [Action taken]
  - Result: [SUCCESS|FAILED|ESCALATED]
- Context Delta:
  - Added: [New knowledge/artifacts]
  - Changed: [Modified state]
- Suggested next agent: [AgentName]
- Next action: [CONTINUE|MERGE|VERIFY|ESCALATE|ABORT]

_STEP_COMPLETE Format

_STEP_COMPLETE:
  Agent: [Name]
  Branch: [branch_id if parallel, else "main"]
  Status: SUCCESS | PARTIAL | BLOCKED | FAILED
  Output:
    type: [Output type]
    summary: [Brief summary]
    files_changed: [List if applicable]
  Handoff:
    Format: [AGENT_TO_AGENT_HANDOFF format]
    Content: [Full handoff for next agent]
  Artifacts:
    - [List of produced artifacts]
  Next: [NextAgent] | VERIFY | DONE
  Reason: [Why this next step]

Handoff Directions (SKILL.md excerpt)

Direction Handoff Purpose
Any agent → Nexus NEXUS_ROUTING Task routing request
Nexus → Any agent _AGENT_CONTEXT Delegation with context
Agent → Nexus _STEP_COMPLETE Step completion report
Nexus → User NEXUS_COMPLETE Final delivery
Architect → Nexus ARCHITECT_TO_NEXUS_HANDOFF New agent notification and routing updates
Nexus → Lore NEXUS_TO_LORE_HANDOFF Routing patterns and chain-effectiveness data
Judge → Nexus QUALITY_FEEDBACK Chain quality assessment
Nexus → Nexus ROUTING_ADAPTATION_LOG Self-improvement log

External feedback sources: Nexus[deliver] (epic-chain results), Judge (quality), Architect (new agents), Lore (validated routing knowledge), Darwin (ecosystem evolution signals).

NEXUS_COMPLETE Required Elements (SKILL.md excerpt)

  • ## NEXUS_COMPLETE header (canonical template: reference/output-formats.md)
  • Task description and acceptance criteria
  • Chain selected and mode used
  • Per-step results with agent, status, and output summary
  • Verification results (tests, build, security) — evidence-bound; unexercised paths labeled UNVERIFIED (Q10)
  • Acceptance Provenance — every intent-contract criterion classified verified/partial/missed/dropped(DEC-n), none silent (Q15)
  • Decision Ledger — DEC-n judgment calls made without the user, interpretation entries first; omit only when empty (Q4-Q6)
  • ## Prompt Tuning trace when any spawn's directives were adapted (field, old→new, trigger, reward_basis), delta-only — omit entirely when no spawn was tuned
  • Summary with overall status
  • Residual Ledger — each leftover as RES-n (class, blocker/owner, marker location, route), bound bidirectionally to any #TODO(agent): left behind, plus the completion-sweep line, which carries two counts — residue and coverage (scanned, 0 hits; 7 changed / 7 evidenced), never omitted

Source: SKILL.md on GitHub

3 warnings13d5 checks · Risk MEDIUM
  • Gen Agent Trust Hub13d

    The 'nexus' skill is a comprehensive multi-agent orchestration framework that manages complex task chains. While it incorporates extensive internal guardrails and verification protocols, it explicitly mandates the use of high-risk flags that bypass security permissions to achieve autonomy. It also provides instructions for establishing persistent tasks via cron and GitHub Actions, and utilizes external research tools to fetch content from the web.

  • Socket13d

    1 alert: gptSecurity

  • Snyk13d

    Risk: LOW · No issues

  • Runlayer6mo

    6/22 files flagged

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at c805268. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/nexus