agent-ecosystem-radar

Agent Ecosystem Radar

Purpose

Use this as the set entrypoint for agent ecosystem discovery. It chains four GroundLine skills into one decision loop:

research-agent-ecosystem -> evaluate-agent-capability -> compare-agent-workflows -> recommend-groundline-upgrades

If the runtime can load named skills, use them in that order. If not, run the same phases inline using the contracts in references/output-contracts.md.

Routing Boundary

Use this when the user asks for research, comparison, and recommendation in one pass. Use the individual skills when the request clearly asks for only one stage: source gathering, single-candidate evaluation, candidate comparison, or upgrade recommendation from existing findings.

Workflow

1. Scope the target: user goal, supported runtime, current GroundLine surface, and whether network research is allowed.
2. Run research-agent-ecosystem to collect source-backed options.
3. Run evaluate-agent-capability for any existing tool, skill, plugin, MCP server, hook, agent, or workflow pack that may be adopted.
4. Run compare-agent-workflows to score overlap, risk, context cost, setup surface, and GroundLine fit.
5. Run recommend-groundline-upgrades to classify each candidate as adopt, adapt, watch, or reject.
6. End with exact next work items and verification gates.

Source Hints

Use references/external-workflow-interop.md for known source families and references/source-registry.json for radar seeds. Prefer official docs and primary repositories. Treat registry listings and blog posts as discovery signals until a primary source confirms the claim.

Safety Rules

• Do not install third-party skills, plugins, hooks, MCP servers, or agents during research.
• Do not enable hooks or mutate provider homes while comparing options.
• Do not recommend broad context loading when a narrow skill or reference file would do.
• Keep unsupported providers out of the GroundLine target scope.

Output Contract

Return all four sections in order:

GroundLine Research:
- scope:
- primary sources:
- secondary sources:
- candidate list:
- confirmed facts:
- unverified claims:
- uncertainty:

GroundLine Capability Evaluation:
- target:
- artifact type:
- source evidence:
- unverified claims:
- context cost:
- setup and mutation surface:
- security risk:
- maintenance signal:
- GroundLine fit:
- decision:
- verification needed:

GroundLine Comparison:
- candidates:
- scoring axes:
- strongest matches:
- overlap with GroundLine:
- context and setup cost:
- rejected or deferred:
- comparison gaps:

GroundLine Recommendation:
- adopt:
- adapt:
- watch:
- reject:
- next GroundLine changes:
- side-effect boundary:
- verification checklist: