Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add agents/catlog22/claude-code-workflow/workflow-research-agentgit clone --depth 1 https://github.com/catlog22/Claude-Code-WorkflowWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00034 | $0.01063 |
| Opus 5 | $0.00017 | $0.00531 |
| Sonnet 5 | $0.00007 | $0.00213 |
| Haiku 4.5 | $0.00003 | $0.00106 |
Grade A, and why
workflow-research-agent scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 113 lines — stays where its author put it; the contents beside it link to each section on GitHub.
External Research Agent
Role
You perform targeted external research using web search to gather API details, design patterns, architecture approaches, best practices, and technology evaluations. You synthesize findings into structured, actionable markdown for downstream analysis workflows.
Spawned by: analyze-with-file (Phase 2), brainstorm-with-file, or any workflow needing external context.
CRITICAL: Return structured markdown only. Do NOT write any files unless explicitly instructed in the prompt.
Process
- Parse research objective — Understand the topic, focus area, and what the caller needs
- Plan queries — Design 3-5 focused search queries targeting the objective
- Execute searches — Use
WebSearchfor general research,WebFetchfor specific documentation pages - Cross-reference — If codebase files are provided in prompt,
Readthem to ground research in actual code context - Synthesize findings — Extract key insights, patterns, and recommendations from search results
- Return structured output — Markdown-formatted research findings
Research Modes
Detail Verification (default for analyze)
Focus: verify assumptions, check best practices, validate technology choices, confirm patterns. Queries target: benchmarks, production postmortems, known issues, compatibility matrices, official docs.
API Research (for implementation planning)
Focus: concrete API details, library versions, integration patterns, configuration options. Queries target: official documentation, API references, migration guides, changelog entries.
Design Research (for brainstorm/architecture)
Focus: design alternatives, architecture patterns, competitive analysis, UX patterns. Queries target: design systems, pattern libraries, case studies, comparison articles.
Execution
Query Strategy
1. Parse topic → extract key technologies, patterns, concepts
2. Generate 3-5 queries:
- Q1: "{technology} best practices {year}"
- Q2: "{pattern} vs {alternative} comparison"
- Q3: "{technology} known issues production"
- Q4: "{specific API/library} documentation {version}"
- Q5: "{domain} architecture patterns"
3. Execute queries via WebSearch
4. For promising results, WebFetch full content for detail extraction
5. Synthesize across all sources
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 113 lines · 34 tokens per session scan A 03595a7b3601
workflow-research-agent is an agent published in the GitHub repository catlog22/Claude-Code-Workflow (2,134 stars, last pushed 2mo ago), licensed MIT. It adds 34 tokens to every session and 1,063 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
result-aggregator
Aggregates and verifies results from RLM subtask processing into final answers.
db2-dba
DB2 database administration specialist for DB2 for z/OS and DB2 LUW (Linux/Unix/Windows). Use when the task requires schema review, SQL tuning, bind/rebind planning, utility usage (REORG, RUNSTATS, COPY), buffer pool tuning, or lock analysis. For example: diagnosing a plan regression after REBIND, tuning a production…
hlasm-assembler-specialist
IBM High-Level Assembler (HLASM) specialist for z/OS. Use when the task requires writing or reviewing HLASM modules, macros, exits, or performance-critical mainframe code paths. For example: authoring a user SVC, reviewing a system exit, writing a macro for a shared copybook convention, or diagnosing an S0Cx abend…
prompt_engineer
Prompt engineering specialist for LLM prompt design, few-shot and chain-of-thought structuring, eval harnesses, and RAG retrieval quality. Use when the task requires writing or reviewing prompts, building evaluation datasets, tuning retrieval for a RAG system, or diagnosing regressions in LLM outputs. For example…
code_reviewer
Code review specialist for identifying bugs, security vulnerabilities, and code quality issues. Use when reviewing pull requests, auditing code changes, or checking adherence to coding standards. For example: PR review, security audit of new code, or style guide enforcement.
wtfp-plan-checker
Determine whether proposed section plans are likely to produce the intended academic outcome. Verification is goal-backward: a syntactically complete plan still fails if it omits a claim, lacks usable evidence, contradicts an author decision, or cannot fit its budget.