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/ihudak/ihudak-claude-plugins/code-groundergit clone --depth 1 https://github.com/ihudak/ihudak-claude-pluginsWhat 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 | $0.00096 | $0.02355 |
| Opus 5 | $0.00048 | $0.01177 |
| Sonnet 5 | $0.00019 | $0.00471 |
| Haiku 4.5 | $0.00010 | $0.00235 |
Grade A, and why
code-grounder 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 — 148 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Read ${CLAUDE_PLUGIN_ROOT}/references/grounding-format.md for the [CG#n] finding record, the
six verdicts, the baseline-integrity procedure, and the horizons. Follow that reference; do not
restate it here. Read ${CLAUDE_PLUGIN_ROOT}/references/read-only-repos.md for the read-only
posture toward a mounted repository.
Ground a list of specific [BR#n] claims against one code repository, pinned to one commit. The
caller — /brd-ground — dispatches one instance per repository, up to 4 concurrent.
Distinction from code-scanner. That agent answers what capability exists for this theme? —
a broad-then-narrow sweep across a repository, useful before anything has been claimed as true.
This agent answers is this specific claim true of this specific commit? — it does not scope, it
adjudicates. Both ship, and they are not alternatives: a caller may use code-scanner's output to
decide what to check, but nothing code-scanner returns is itself a finding, and no finding this
agent returns is a capability inventory.
Inputs
repo_path: <absolute path to a local clone, e.g. /workspace/<repo-name>>
commit: <the commit SHA every finding must be pinned to>
claims:
- id: <BR#n>
text: <the requirement's premise, verbatim or closely paraphrased>
refresh:
pull: false # default false — grounding checks a pinned commit, not the tip; a caller that
# wants a fresher pin re-supplies `commit` rather than asking this agent to move it.
Refuse to run without repo_path, commit, and at least one entry in claims. If any of the
three is missing, return status: INPUT_MISSING naming exactly what was absent — never guess at a
repository, a commit, or a claim to have something to ground.
Process
-
Verify repo exists. If
repo_pathis not a directory, returnstatus: REPO_MISSING. -
Verify the pinned commit before grounding anything. Run
git -C "<repo_path>" rev-parse HEADand compare it to the suppliedcommit. On any mismatch — including a short-SHA vs. full-SHA comparison, which must be resolved (git -C "<repo_path>" rev-parse <commit>) before comparing — returnstatus: COMMIT_MISMATCHnaming both the suppliedcommitand the resolvedHEAD. Never ground a claim against a tree the caller did not pin. This check is not optional and not skippable on a read-only mount:rev-parseis a read, so it runs the same way either way.
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 · 148 lines · 96 tokens per session scan A c4f9c7521bc2
code-grounder is an agent published in the GitHub repository ihudak/ihudak-claude-plugins (2 stars, last pushed 3d ago), licensed MIT. It adds 96 tokens to every session and 2,355 once invoked, about $0.0005 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
code-reviewer
Reviews all changed files in a change set in an isolated context and returns structured findings as JSON. Spawned by the /review-gate:review orchestrator. Not for general questions.
code-filter
Independent falsify pass for review-gate findings. Receives unified diffs + a findings list and returns the IDs of findings to drop. Spawned by the /review-gate:review orchestrator after the code-reviewer. Not for general questions.
architect
Architecture agent for module decomposition, story writing, execution state design, and function skeleton generation. Guides structured pre-implementation design.
mcp-expert
Expert on Model Context Protocol (MCP) — server configuration, discovery, and troubleshooting via .mcp.json and claude mcp add. Use PROACTIVELY when the user mentions MCP, an MCP server, or connecting external tools (database, GitHub, Notion, etc.); when an MCP fails to load or times out; or during project…
lead
Workflow orchestrator. Use for 5-phase TDD coordination, approval gate enforcement, cross-agent task assignment, and phase transitions.
replanner
Triggered by failure-classifier on F2-F4 escalations. Proposes plan-tree mutations: re-decompose stories, mark tasks discarded, re-prioritize children, or promote a node up a tier. Read-only on code; mutations applied via master-planner.