- Home
- Skills
- Agents & Orchestration
- Token Governor
Works with the AI Agents you already use
Token Governor
Keep long or broad agent tasks focused with a locate-inspect-validate-broaden ladder, a KEEP/DROP/DEFER context ledger, and a clear next minimum check.
$14
Token Governor
Example session with this skill installed
This large repo task is using too much context. Keep the session lean. Start with the acceptance criteria and failure trace, inspect only the relevant files, preserve facts needed for the next step, and give me KEEP, DROP, DEFER, EXPENSIVE FACTS PRESERVED, and NEXT MINIMUM READ/CHECK. Do not skip a required check or assume a tool exists.
- Read your context and instructions
- Compiled the token governor
Illustrative example (synthetic; no repository was inspected).
KEEP
- the failing test name and relevant stack frame
- the two files on the suspected path
- the ADR that defines the invariant
DROP
- full file dumps already reviewed
- verbose passing-test logs
- the initial broad directory listing
DEFER
- browser or end-to-end checks until a code-level hypothesis exists
EXPENSIVE FACTS PRESERVED
- the reproduction command
- any supplied commit or bisect result, recorded only if it is present in the evidence
NEXT MINIMUM READ/CHECK
- inspect only the suspected function and its relevant contract
- run the narrowest authorized check that confirms or refutes the hypothesis; if the needed tool or evidence is unavailable, mark the result UNVERIFIED or BLOCKED
This follows the package's example output shape for a synthetic debugging scenario. It does not report an executed test, verified commit or actual repository finding.
Connects securely to your tools. The creator never sees your data.
About this skill
Token Governor is for long or broad tasks where repeated file reads and oversized tool output are consuming context. It uses a locate, inspect, validate, broaden ladder, then returns a KEEP, DROP, DEFER ledger with expensive facts preserved and the next minimum read or check.
Use it when context overhead matters or when you explicitly ask to keep a session lean. It starts with metadata and exact files, inspects the smallest relevant evidence, performs a focused validation when the required capability is available, and broadens only when risk or a failure warrants it. The context ledger keeps expensive facts while repeated dumps, redundant logs and speculative branches can be dropped or deferred.
The workflow gives project governance, accepted contracts and current evidence higher authority than a generic preference or this skill. It verifies tool availability and labels unavailable evidence UNVERIFIED or BLOCKED. Saving context never overrides safety, acceptance criteria, provenance or critical-path validation.
The output has five parts: KEEP, DROP, DEFER, EXPENSIVE FACTS PRESERVED, and NEXT MINIMUM READ/CHECK. Compared with a plain recap, it also names the smallest next evidence check and the facts that need to survive a handoff.
It is not intended for a small task whose working set is already minimal. The package does not guarantee model output or host-specific behavior. Dependency installation, commits, pushes, merges, deployments and destructive migrations are never implied; project-specific authority and explicit authorization remain in control.
It is intended for long or broad tasks where context overhead matters, not small tasks with a minimal working set. The package does not guarantee model output, measured token savings or host-specific behavior. If required evidence or a capability is unavailable, the workflow marks that limit rather than inventing results.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 17 days ago
Permissions
Read Files is required. Terminal/Shell and Write Files are optional enhancements with documented fallbacks; Browser, Network, MCP and environment variables are not required.