All skills
simota avatar

/architect

@965f4f9
by shingo imotasimota/agent-skills85 stars
15

Designing new skill agents via gap analysis, overlap detection, SKILL.md + reference generation, and Nexus integration. Not for task orchestration (Nexus) or format-only audits (Gauge).

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

This session only. Nothing lands on disk.

referencenexus-integration.md

≈549 tokens on demand. Your agent reads this file only when SKILL.md points to it.

New-Skill Nexus Integration

Use when adding or changing a routing surface. The shared envelopes and hub behavior are authoritative in _common/AUTORUN.md and _common/HANDOFF.md; read them rather than reimplementing _AGENT_CONTEXT, NEXUS_HANDOFF or direct-spawn instructions here. The skill-specific completion fields remain in reference/autorun-schema.md.

Routing Records

Choose an existing task type when it accurately represents the owned outcome; add a type only for a real uncovered route. Inspect the current routing registry, not a cached list of task types. Record both the chain placement and positive/negative triggers:

ROUTING_ENTRY:
  task_type: "[TASK_TYPE]"
  simple_chain:
    - "[Agent1]"
    - "[NewAgent]"
    - "[Agent2]"
  complex_chain:
    - "[Agent1]"
    - "[Sherpa]"
    - "[NewAgent]"
    - "[Agent2]"
    - "[Agent3]"
  additions:
    - condition: "[Condition]"
      add: "[Agent]"
TRIGGER_REGISTRATION:
  keywords:
    - "[keyword1]"
    - "[keyword2]"
    - "[keyword3]"

  patterns:
    - "[pattern with wildcards]"

  negative_patterns:
    - "[patterns that should NOT trigger]"

Integration Gate

  • A direct positive request and a Nexus-chain request both reach the intended owner; negative requests stay with their existing owner.
  • Upstream inputs, downstream output fields and partial/blocked behavior agree with the actual partner contracts.
  • The hub receives pending confirmations, completion state and a proposed next step; a handoff recommendation never directly invokes another specialist.
  • Shared paths resolve from the installed skill directory and required contracts are actually delivered. Optional project-local partners follow _common/PROJECT_LOCAL_SKILLS.md.
  • Update the current routing/roster registries required by the repository's authoring instructions; do not maintain copies in this reference.
  • Exercise new/changed routing and error cases. Record unexecuted cases as unverified.

Use the existing hub guardrails and owning lifecycle rules for escalation or rollback. A failed integration test does not authorize deleting an existing skill or bypassing its retirement/migration approvals.

Source: SKILL.md on GitHub

2 warnings13d5 checks · Risk MEDIUM
  • Gen Agent Trust Hub13d

    The skill features a self-evolution capability that allows it to autonomously modify its own reference files and instructions. It also acts as an agent design tool, which introduces a surface for indirect prompt injection based on untrusted user requirements.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

  • Runlayer6mo

    5/18 files flagged

  • ZeroLeaks5mo

    1 finding · Score: 69/100

Signed by skilld at 965f4f9. 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/architect