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.
git clone --depth 1 https://github.com/harnessworks/harness-starter-kitWrote 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/harnessworks/harness-starter-kit/harness-refresh)<a href="https://agentmods.dev/commands/harnessworks/harness-starter-kit/harness-refresh"><img src="https://agentmods.dev/badge/commands/harnessworks/harness-starter-kit/harness-refresh/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/commands/harnessworks/harness-starter-kit/harness-refresh"><img src="https://agentmods.dev/badge/commands/harnessworks/harness-starter-kit/harness-refresh.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00000 | $0.00999 |
| Opus 5.5 | $0.00000 | $0.00400 |
| Sonnet 5.5 | $0.00000 | $0.00200 |
| Haiku 4.5 | $0.00000 | $0.00100 |
Grade A, and why
harness-refresh 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 2mo 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 — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/harness refresh
Refresh a target repository's existing harness by reviewing whether its agent-facing rules, knowledge store, and drift checks are still accurate.
Harness Refresh is maintenance for the target harness. It does not refresh the
local ./harness-starter-kit reference clone and does not replace
/harness update.
Goal
Inspect the target repository's current harness, classify stale or conflicting artifacts, apply only safe documentation/check updates, and finish with a clear Harness Refresh Report.
Use this command when the maintainer wants to reduce old guidance, duplicate records, obsolete failure notes, stale decisions, or unused harness checks.
Procedure
- Treat the current working directory as the target repository root.
- Inspect target repository state before changing files:
git status --short --branchAGENTS.md,CLAUDE.md, README files, contribution docs, and CI configs.harness/source.json, if presentdocs/decisions/,docs/failures/,docs/conventions/, anddocs/domain/scripts/check_*.py,scripts/check_harness.py, package scripts, test configs, lint configs, and pre-commit config
- Build a short inventory of harness artifacts and classify each relevant item:
keep: still accurate and useful.update: useful but stale, incomplete, or unclear.merge: duplicate or overlapping guidance that should be consolidated.archive/delete candidate: likely obsolete or unused, but not removed without explicit user approval.manual review: needs a maintainer decision because it changes workflow, CI, dependencies, architecture rules, or historical records.
- Check for:
- stale docs that reference missing commands, paths, or tools
- duplicated guidance split across
AGENTS.md, README files, and docs - obsolete decisions or failure records contradicted by current code
- harness checks that no normal command, CI workflow, or documentation uses
- deterministic, local, non-network, reasonably fast checks for product behavior that agents are expected to verify repeatedly, but are not included in the documented normal completion gate and have no recorded reason for remaining focused or manual
- rules that describe architecture constraints no check or review point can enforce
- For large docs directories, list files first and read only records relevant to the refresh question, stale/conflict signals, or current harness rules.
- Patch only safe, target-specific updates that preserve the repository's current architecture, tools, commands, and documentation style.
- Prefer marking decision and failure records as superseded or obsolete over deleting them. If a replacement record is clearer, add or update the replacement and cross-reference the old record.
- Do not delete, archive, move, or wholesale replace files unless the user explicitly approves the specific files after seeing the candidates.
- Run relevant local checks, such as docs drift, structure drift,
effectiveness plan checks, project tests, linting, type checks, or
/harness doctor.
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.
- 2mo ago First seen · 112 lines · 0 tokens per session scan A ae51a788226d
harness-refresh is a command published in the GitHub repository harnessworks/harness-starter-kit (115 stars, last pushed 4mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 999 tokens. 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 commands, from other repositories
latch-compact
Manually trigger a latch compaction for the current session.
latch-decay
Apply weekly decay + promote staging nodes that cleared the refcount bar.
remember
Persist one explicit user-originated project memory.
recall
Recall durable project context for a concrete question.
memory-gc
Garbage collection for stale memory entries - identify and clean up obsolete content.
memory-review
Display current memory state with timestamps, sizes, and staleness indicators.