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/synaptiai/agent-capability-standard/constrainnpx skills add synaptiai/agent-capability-standard --skill constraingit clone --depth 1 https://github.com/synaptiai/agent-capability-standardWrote 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/synaptiai/agent-capability-standard/constrain)<a href="https://agentmods.dev/skills/synaptiai/agent-capability-standard/constrain"><img src="https://agentmods.dev/badge/skills/synaptiai/agent-capability-standard/constrain.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 | $0.00039 | $0.02258 |
| Opus 5 | $0.00019 | $0.01129 |
| Sonnet 5 | $0.00008 | $0.00452 |
| Haiku 4.5 | $0.00004 | $0.00226 |
Grade A, and why
constrain 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 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.
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 — 311 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Intent
Execute constrain to evaluate proposed actions against defined policies, enforce permission boundaries, and reduce scope to the minimum required for safe operation. This is the gatekeeper that prevents unsafe actions before they execute.
Success criteria:
- All relevant policies evaluated for the proposed action
- Clear allow/deny decision with rationale
- Scope reduced to minimum necessary permissions
- Violation actions documented (block, warn, log)
Compatible schemas:
schemas/output_schema.yaml
Inputs
| Parameter | Required | Type | Description |
|---|---|---|---|
action |
Yes | object | The proposed action to evaluate (capability, target, parameters) |
policy_id |
No | string | Specific policy to evaluate against (default: all applicable) |
scope |
No | object | Current permission scope to constrain |
enforcement_level |
No | enum | strict (block), advisory (warn), audit (log only) |
context |
No | object | Additional context for policy evaluation (user, session, etc.) |
Procedure
-
Identify applicable policies: Find all policies relevant to the action
- Load policy definitions from
.policies/or policy registry - Match action type to policy triggers
- Consider context (user role, environment, time) for policy selection
- Load policy definitions from
-
Evaluate preconditions: Check if action meets basic safety requirements
- Is action within workspace boundaries?
- Does action target sensitive resources (.env, credentials)?
- Is checkpoint required but missing?
- Is approval required but not obtained?
-
Apply policy rules: Evaluate each policy condition
- Parse policy conditions (AND/OR/NOT expressions)
- Check each condition against action parameters
- Record which conditions pass and which fail
- Determine overall policy verdict (allow/deny/conditional)
-
Reduce scope: Apply least privilege principle
- Identify minimum permissions needed for action
- Remove unnecessary permissions from scope
- Document removed permissions and rationale
What ships with it
3 files 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 · 311 lines · 39 tokens per session scan A 5683cba731de
constrain is a skill published in the GitHub repository synaptiai/agent-capability-standard (4 stars, last pushed 6d ago), licensed Apache-2.0. It adds 39 tokens to every session and 2,258 once invoked, about $0.0002 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 skills, from other repositories
token
Use when the user wants to plant a decoy/tripwire secret, create a canary token, or check whether a planted token has been leaked to Claude.
gh
GitHub API access and project management automation for the hallucination-detector repo. Uses octokit with proxy-aware client for all GitHub operations — issues, PRs, labels, milestones, Projects V2. No gh CLI required.
scan
Use when the user worries they pasted a secret, asks "did I leak anything?", or wants a full-history privacy audit — the automatic scan only covers the latest message.
evaluate-options
Research and evaluate implementation options before presenting a recommendation. Use when multiple approaches exist for a problem and a decision is needed. Launches one background research agent per option in parallel, collects evidence-backed findings, then presents a recommendation grounded in observed data — not…
audit
Use when the user wants to check installed Claude Code plugins, skills, agents, or MCP configs for leaked secrets or suspicious/exfiltration patterns.
leaked
Use when the user asks what PII they've leaked, wants the exposure dashboard/stats, or wants to reset the counter.