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 instructions/ksegit/version-sentinel/gemini-mdgit clone --depth 1 https://github.com/KSEGIT/Version-SentinelWhat 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 | $0.00506 | $0.00506 |
| Opus 5 | $0.00253 | $0.00253 |
| Sonnet 5 | $0.00101 | $0.00101 |
| Haiku 4.5 | $0.00051 | $0.00051 |
Grade A, and why
Version-Sentinel GEMINI.md scanned grade A with 1 finding 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.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
`bash`, `jq`, `curl`, and `python3` (3.11+, for `tomllib`) on `PATH`. What it actually says
version-sentinel (Gemini CLI extension)
This extension hard-blocks dependency additions, bumps, and downgrades until a fresh, source-cited version check is recorded. It exists to stop the model from shipping a hallucinated or stale package version remembered from training data.
How it works
- You try to edit a dependency manifest (
write_file/replaceonpackage.json,requirements*.txt,pyproject.toml,Cargo.toml,*.csproj, ...) or run an install command viarun_shell_command(npm install,pip install,cargo add,dotnet add package, ...). - A
BeforeToolhook fires and exits 2 — the tool call is blocked and the hook's stderr explains why:BLOCKED: version-sentinel. Package: lodash (npm). Version: 4.17.21. No fresh version check on record. - To unblock: look up the real latest version on the upstream registry
(npmjs.com, pypi.org, crates.io, nuget.org), then record the check:
Then retry the original edit or install — the hook finds the fresh entry and lets it through./vs-record npm lodash 4.17.21 https://www.npmjs.com/package/lodash
Intentional pins
Pinning an old version on purpose is fine — record it with a reason instead of a URL:
/vs-record npm pkg 1.0.0 "intentional: CVE fix deferred pending audit"
Intentional pins unblock the hook and show as intentional-pin (not DRIFT)
in audits.
Auditing drift
Run /check-versions to scan manifests within 4 directory levels of the workspace
and compare each pinned version against the latest upstream release. Advisory only
— never blocks.
Escape hatch
Set VS_DISABLE=1 in the environment to make every version-sentinel hook a
no-op (useful for throwaway sessions).
Prerequisites
bash, jq, curl, and python3 (3.11+, for tomllib) on PATH.
Recorded checks live in <workspace>/.version-sentinel/checks.json
(auto-gitignored on first write).
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 · 57 lines · 506 tokens per session scan A 1c623ab23ab9
Version-Sentinel GEMINI.md is an instructions file published in the GitHub repository KSEGIT/Version-Sentinel (3 stars, last pushed 2d ago), licensed MIT. It adds 506 tokens to every session, about $0.0025 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
ai-dev-kit AGENTS.md
Instructions for noah-sheldon/ai-dev-kit, covering ai dev kit — agent instructions, core principles, contributor policy, available agents and agent orchestration.
magi-workflow CLAUDE.md
Instructions for howar31/magi-workflow, covering what this is, slash commands, subagents, project state model and project document tiers.
Cerebro AGENTS.md
Instructions for AgenticFirst/Cerebro, covering cerebro — agent context, read before touching integrations, project layout, conventions a fresh agent would otherwise miss and running locally.
discoclaw CLAUDE.md
Instructions for DiscoClaw/discoclaw, covering discoclaw, safety, context loading (strict), identity and workspace file ownership.
nextjs-skills CLAUDE.md
Instructions for mohamed-hossam1/nextjs-skills: All contribution rules — style, naming, file layout, verification — live in AGENTS.md. Read that first.
claude-plugins CLAUDE.md
Claude Code instructions for aiocean/claude-plugins, covering claude.md, what this is, plugin structure, marketplace registry and plugin categories.