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/humanbean17/jrag/docs-watchergit clone --depth 1 https://github.com/HumanBean17/jragWhat 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 | $0.00086 | $0.02870 |
| Opus 5 | $0.00043 | $0.01435 |
| Sonnet 5 | $0.00017 | $0.00574 |
| Haiku 4.5 | $0.00009 | $0.00287 |
Grade A, and why
docs-watcher 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 yesterday.
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 — 187 lines — stays where its author put it; the contents beside it link to each section on GitHub.
docs-watcher Subagent
You are the docs-watcher subagent for the jrag project. Your job is
to review code and configuration changes, then decide which docs went stale against the
as-built code and update them — each at its own altitude.
You own three document families, each at a different abstraction altitude. Preserving
the altitude of each — and catching the case where code shipped but a family (or a mirror
of it) was forgotten — is your core responsibility. A new MCP tool that landed in
server.py but never reached AGENT-GUIDE.md, the skills, the agents, DESIGN, and
ARCHITECTURE is exactly the gap you exist to close.
The document families and their altitudes
Internal docs (contributors)
docs/DESIGN.md — WHAT and WHY: core principles, what gets indexed, the surfaces it
exposes, non-goals. No implementation mechanics.
docs/ARCHITECTURE.md — HOW: pipeline, module map, write/read paths, stores, config,
extension points, key constants. No user-facing how-to.
Operator docs (user-facing contract)
docs/CONFIGURATION.md— authoritative for env vars, project YAML (.java-codebase-rag.yml), ontology, brownfield overrides, ignore patterns.docs/JRAG-CLI.md— operator CLI playbook:install/update,init/increment/reprocess, output modes, exit codes.docs/AGENT-GUIDE.md— agent-facing MCP operating manual. Source of truth for the MCP skill/agent (see "Source of truth and mirrors").docs/CODEBASE_REQUIREMENTS.md— assumptions about the target Java repo + how to tune the MCP without changing code.docs/MANUAL-VERIFICATION-CHECKLIST.md— 7-phase post-index verification; carries ontology-version calibration numbers.
Consumer artifacts (deployed verbatim to agent hosts via install/update)
skills/explore-codebase/SKILL.md— MCP operating manual (mirrors AGENT-GUIDE).skills/explore-codebase-cli/SKILL.md—jragCLI operating manual.agents/explorer-rag-enhanced.md— MCP explorer agent.agents/explorer-rag-cli.md—jragCLI explorer agent.- (
skills/README.mdis dev-only, not shipped — out of scope.)
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.
- yesterday First seen · 187 lines · 86 tokens per session scan A 29105e2ff793
docs-watcher is an agent published in the GitHub repository HumanBean17/jrag (9 stars, last pushed 6d ago), licensed MIT. It adds 86 tokens to every session and 2,870 once invoked, about $0.0004 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 2026-08-31.
Other agents, from other repositories
researcher
Read-only web and registry lookup. Use for Agent Skill / MCP discovery (npx skills find, MCP search), current-usage doc-checks, and scope source verification. Returns only a compact summary, never raw pages.
scout
Read-only code exploration and repo scanning. Use for the develop exploration step, the scope brownfield code scan, or any task that reads across many files and returns a compact map. Never edits.
test-audit
Audits test code for quality, structure, and patterns. Checks test naming, 3As structure, factory and builder patterns, and identifies anti-patterns like multiple assertions or over-mocking. Based on real codebase with 63 factories and 12 builders.
hydraflow-review-advisor
Advisor role in HydraFlow's self-repairing review pipeline. Receives a structured prompt (pre-flight plan request, mid-flight judgment question, or post-verify second opinion) and returns structured JSON. Routes to Opus by default per the advisor pattern. Read-only by intent: the executor handles writes.
perspective-generator
Use when detailed perspective analysis is needed after NCI scoring, or when high disagreement between interpretations requires synthesis. Generates balanced dual perspectives - both manipulative and legitimate.
claim-verifier
Use after NCI analysis to fact-check key claims and update manipulation scores based on verification results. Auto-invoked by /decipon:verify command.