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 skills/gldywn/gatekeeper/install-gatekeepernpx skills add Gldywn/gatekeeper --skill install-gatekeepergit clone --depth 1 https://github.com/Gldywn/gatekeeperWrote 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/skills/gldywn/gatekeeper/install-gatekeeper)<a href="https://agentmods.dev/skills/gldywn/gatekeeper/install-gatekeeper"><img src="https://agentmods.dev/badge/skills/gldywn/gatekeeper/install-gatekeeper.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 5.1 | $0.00110 | $0.03982 |
| Opus 5 | $0.00055 | $0.01991 |
| Sonnet 5 | $0.00022 | $0.00796 |
| Haiku 4.5 | $0.00011 | $0.00398 |
Grade B, and why
install-gatekeeper scanned grade B with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
**Codex CLI**, in `~/.codex/config.toml`: Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
version="$(curl -sS https://api.github.com/repos/Gldywn/gatekeeper/releases/latest | grep '"tag_name"' | head -1 | sed 's/.*"v\{0,1\}\([0-9][^"]*\)".*/\1/')" How it starts
The opening of the file, as written. The whole thing — 217 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Install Gatekeeper
Gatekeeper is a human-approved bridge between an agent and a live database: the agent proposes SQL, a human approves it in Beekeeper Studio, and the rows come back. Nothing runs unapproved, and the agent never holds the credentials.
Installing it means putting four pieces in place on one machine. Do every piece your environment lets you do, verify each against something you can observe, and finish with a short, exact list of what only the human can do.
| Piece | What it is | Who does it |
|---|---|---|
| Beekeeper Studio plugin | The approval UI, running inside Beekeeper | You download and extract it; the human restarts Beekeeper |
| MCP server | @gldywn/gatekeeper-mcp-server, launched by the agent's MCP client |
You register it; it only comes alive in a fresh agent session |
gatekeeper skill |
Teaches an agent to drive the tools well | You install it |
| Pairing | A 6-digit code that hands the plugin its capability token | The human types the code, once per machine |
How to run this install
- Never fake a step. Every step below ends with a check you can run. If the check fails or the step is blocked (no network, a sandbox, a refused permission, a path you cannot write), stop that step, keep going with the ones that do not depend on it, and record the blocked one for the handoff at the end with the exact command and what success looks like. A half-done install reported as done is worse than an honest checklist.
- Never work around a refusal. If the harness or the human declines a command, hand it over instead of rewording it.
- Use the paths in this file verbatim. Do not guess a plugins folder or a config location.
- Merge configs, never overwrite them. The MCP client config almost always holds other servers. Read it, add the one key, write it back, then read it again and confirm it still parses and the other entries survived.
- Name the platform up front. Step 0 has the one line to say for macOS, Linux or Windows. Say it before anything else; nobody should discover their platform is untested three steps in.
- When something misbehaves, read TROUBLESHOOTING.md, next to this file. It covers the plugin that never appears in Beekeeper, the server that will not start, pairing codes that come back refused, and approval prompts on every poll. Only read it when you hit one of those; the normal path never needs it.
- Install for the agent you are running in, unless the human names another one. If you cannot tell which harness you are in, ask rather than assume; step 2 depends on the answer.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 5d ago First seen · 217 lines · 110 tokens per session scan B e8a78f31713f
install-gatekeeper is a skill published in the GitHub repository Gldywn/gatekeeper (2 stars, last pushed 3d ago), licensed MIT. It adds 110 tokens to every session and 3,982 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it B with 2 findings (reads agent configuration directories, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
release-checklist
Pre-release safety audit for the Bifrost repo. Scans database migrations changed in a release for high-scale deadlock / lock-contention risks and for work that blocks application boot time, then produces a pass/warn/fail report with a concrete remediation plan. Invoked with /release-checklist [git-ref-range]. Built to…
rls-policy
Write and test Supabase/Postgres Row-Level Security policies so users can only access their own rows. Use when a table holds user data, when enabling RLS, or when auditing access control on Supabase.
db-backup-restore
Safely back up and restore a Postgres/Supabase database — pgdump/pgrestore, Supabase backups/PITR, and a no-data-loss restore runbook. Use before risky migrations, for disaster recovery, or when cloning prod to staging.
migration-safety
Generate safe, reversible Postgres/Supabase migrations using the expand→backfill→contract pattern, avoiding locks and data loss. Use when altering schema, adding/removing columns, or changing types on tables that already hold data.
query-optimizer
Diagnose and fix slow Postgres queries using EXPLAIN ANALYZE, index suggestions, and N+1 detection. Use when a query is slow, a page is slow because of the DB, or before shipping a query that touches a large table.
realtime-subscriptions
Add Supabase Realtime to a Next.js app — Postgres Changes, Broadcast, and Presence — with proper cleanup and RLS-aware channels. Use when building live updates, presence indicators, or collaborative features.