pantyuhov9-web/agentic-workflow-template / .claude/skills/repo-system-research
repo-system-research
VerifiedResearch the open web for enhancement opportunities (Layer 3 — Opus subagent with WebSearch + WebFetch). Topic-driven. Returns structured findings with hype tags + source URLs. Does NOT auto-implement.
System Research — entry point (Layer 3)
This skill is the invocation handle for outward-facing R&D. The architecture, 8 finding categories, anti-hallucination controls, verbatim Layer 3 subagent prompt, topic auto-discovery scoring heuristic, and output rendering format all live in the workflow doc — this file just routes you there.
When this fires
Auto-triggers when the user wants to scan the open web for enhancement opportunities — or explicit /repo-system-research [topic]. Designed for "leftover credits" use, weekly cadence: find one improvement to consider, discuss it, done.
This is Layer 3 of the three-layer system_audit pipeline. Layers 1+2 (audit skill) look INWARD for rot; Layer 3 looks OUTWARD for enhancements. Convergent vs divergent — different jobs.
Arguments
$ARGUMENTS is an optional research topic. If provided, must be ≥3 words (the workflow's Step 3 sanity-checks this). If absent, the auto-discovery heuristic in the workflow proposes 2–3 candidate topics for the user to pick from.
Begin
Read workflows/system_research.md — authoritative spec. The "Orchestration sequence" section defines the 7 steps and contains:
- The verbatim Layer 3 prompt to pass to the subagent (Step 4)
- The topic auto-discovery scoring heuristic (Step 3, mode b)
- The output rendering format (Step 5)
The "Anti-hallucination + anti-hype controls" section enumerates the 8 hard rules (mandatory source URLs, 4-tier authority model, system-map cross-check, additive-only, freshness window, topic anchor, hard search budget).
Then execute Steps 1–7 from the workflow. Hard rule: only ONE enhancement adopted per invocation.
See also
- tools/system_audit.py — Layer 1, runs in Step 1 prelude
- .claude/skills/audit/SKILL.md — Layers 1+2, the convergent counterpart
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |