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 commands/akahkhanna/groundtruth/groundtruth-helpgit clone --depth 1 https://github.com/akahkhanna/groundtruthWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/akahkhanna/groundtruth/groundtruth-help)<a href="https://agentmods.dev/commands/akahkhanna/groundtruth/groundtruth-help"><img src="https://agentmods.dev/badge/commands/akahkhanna/groundtruth/groundtruth-help.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.00008 | $0.00608 |
| Opus 5 | $0.00004 | $0.00304 |
| Sonnet 5 | $0.00002 | $0.00122 |
| Haiku 4.5 | $0.00001 | $0.00061 |
Grade A, and why
groundtruth-help 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 4d 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.
What it actually says
Explain to the user, concisely:
Groundtruth audits whether the agent actually did what it was asked — grounding every claim against the diff / tool-ledger / disk, never the agent's self-report.
It runs automatically every turn (the Stop hook) and checks:
- Honesty & completeness — the claims contract (v2 default). The agent ends each code-changing turn with one fenced
groundtruth-claimsmanifest; Groundtruth verifies it against reality: CA (claimed-but-absent) — a claim the diff/transcript don't support (atests_passthat never ran or ran red, a file not in the diff, a symbol not defined); UC (undeclared change) — a changed file no claim covers; NC (no contract) — a missing/invalid manifest on a code turn. Declareddeferreditems surface as the agent's own set-aside. Opt out withGROUNDTRUTH_CONTRACT=0. - Rules — hardcoded secrets, new tables without RLS, permissive
USING(true)policies, and project rules compiled from your docs (CLAUDE.md / skills) that were in context but broken anyway. - Code debt — stub/placeholder markers, phantom imports, dropped-symbol dangling refs (a "refactor, everything preserved" that left a caller pointing at nothing);
/groundtruth-auditinventories it repo-wide.
A verdict card is written to .claude/groundtruth/<session>.md each turn. With GROUNDTRUTH_BLOCK=1 a fixable block-severity catch (a contract CA; a built-in secret/RLS/.env; or an NC — no manifest — once the repo is contract-aware, i.e. /groundtruth-setup has written the instruction) halts the stop, hands back a corrective (for NC, the schema itself), and re-checks (retry cap 2, then escalates — never wedges); otherwise it's warn-only.
Commands:
/groundtruth— show the latest verdict card/groundtruth-audit— whole-repo deterministic debt scan/groundtruth-rules— review + approve the rules compiled from your docs (the permission gate)/groundtruth-rules-ai— opt-in model pass that proposes richer rules (routed through the same gate)/groundtruth-block on｜off— turn block mode on/off (default: warn)/groundtruth-setup— one-shot installer: writes the contract instruction, arms clean rules, badge/env/groundtruth-help— this
The honest ceiling: Groundtruth verifies the agent did what it said, not that what it said was right. Mechanics, not semantics — correctness still needs a test or a judge.
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.
- 4d ago First seen · 26 lines · 8 tokens per session scan A 0adedf1c906d
groundtruth-help is a command published in the GitHub repository akahkhanna/groundtruth (6 stars, last pushed 1mo ago), licensed MIT. It adds 8 tokens to every session and 608 once invoked, about $0.0000 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 commands, from other repositories
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.
implement
Execute the implementation plan by processing and executing all tasks defined in tasks.md.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.