agentleFS
Ingramradical235Ingramradical235/anty-framework/skills/critical-chain/SKILL.md

critical-chain

Skill1 starsChanged 8 months agoMITView on GitHub

Theory of Constraints 5 Focusing Steps, buffer management with 3-zone traffic light, relay runner execution discipline, WIP limits, and late start scheduling. Use when managing Goal execution pace, detecting bottlenecks, or when buffer status changes.

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When the founder asks "are we on track?"

    line 14
  2. Distinctive:

    SUBORDINATE: All other Drivers serve the constraint.

    line 24
  3. Distinctive:

    ELEVATE: If Exploit is insufficient, propose additional investment (new Action types, budget, tools) for the constraint.

    line 25
  4. Distinctive:

    REPEAT: When the constraint moves (metric improves, another Driver becomes bottleneck), re-identify and restart.

    line 26

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Critical Chain & Buffer Management

When to Apply

  • Managing Goal execution pace and timeline
  • Buffer zone changes (GREEN/YELLOW/RED)
  • Detecting and resolving bottlenecks
  • Prioritizing which Actions to execute next
  • When the founder asks "are we on track?"

Core Framework

TOC 5 Focusing Steps

Applied to the KPI tree. The Crux Driver (highest importance x tractability) is the constraint:

  1. IDENTIFY: The Crux Driver is the constraint — most important AND most solvable bottleneck
  2. EXPLOIT: Maximize that Driver's throughput. Execute its Actions first, eliminate waste, remove blockers. Never let constraint Actions wait.
  3. SUBORDINATE: All other Drivers serve the constraint. Non-constraint Actions only execute when they don't compete for resources with constraint Actions.
  4. ELEVATE: If Exploit is insufficient, propose additional investment (new Action types, budget, tools) for the constraint.
  5. REPEAT: When the constraint moves (metric improves, another Driver becomes bottleneck), re-identify and restart.

Buffer Management (3-Zone Traffic Light)

Buffer sizing: Project Buffer = 50% of estimated time to achieve Goal.

Feeding buffers: Each non-constraint Driver has a small feeding buffer protecting the constraint. If a non-constraint Action delays, the feeding buffer absorbs it.

ZoneBuffer ConsumedAgent Behavior
GREEN (0-33%)On trackNormal execution. Continue planned Actions. Light monitoring.
YELLOW (33-67%)Falling behindInvestigate root cause. Generate recovery Actions. Notify user. Increase scan frequency.
RED (67-100%)Significantly behindAuto-propose high-impact recovery. Escalate to user. Deprioritize low-impact Drivers. Consider scope adjustment.
Buffer consumed: 45%  [========----------]
Goal progress:   34%  [======------------]
Status: YELLOW — buffer consumption outpacing progress

Buffer as early warning: Unlike milestone tracking (detects problems after deadline miss), buffer management provides continuous visibility. YELLOW at 20% progress gives weeks of lead time.

Relay Runner Principle

Three rules for Action execution:

  1. Start immediately — When approved, execute in the very next chain invocation. No delay. (Prevents Student Syndrome — postponing when there's perceived slack.)
  2. Full focus, one at a time — Constraint Driver Actions execute serially. Work one Action to completion before starting the next. (Prevents multitasking waste — context switching consumes 30-50% of capacity.)
  3. Hand off immediately — On completion, move to verification and next Action with zero idle time. Early completion is celebrated. (Prevents Parkinson's Law — work expanding to fill time.)

WIP Limits

ScopeMax In Progress
Constraint Driver1 Action at a time
Non-constraint Drivers3 Actions concurrently
Total system5 Actions across all Drivers

Late Start Scheduling

Non-constraint Driver Actions start as late as possible (while respecting feeding buffers). This reduces unnecessary WIP and keeps focus on the constraint.

Decision Rules

  1. Constraint Actions never wait — always first priority, immediate execution
  2. Buffer zone determines response intensity — GREEN: cruise. YELLOW: investigate. RED: escalate.
  3. WIP limits are hard limits — do not exceed, even if capacity seems available
  4. Relay runner is non-negotiable — no batching of constraint Actions, no multitasking
  5. Re-identify constraint after every metric update — the bottleneck can shift
  6. Subordination means saying no — non-constraint work that threatens constraint throughput is blocked

Anti-Patterns to Detect

Anti-PatternSignalResponse
Multitasking on constraintMultiple constraint Actions in parallel"Constraint Driver must have exactly 1 Action in progress. Serialize."
Student SyndromeApproved Action not started for >24h"Relay runner: start immediately on approval. No delay."
Ignoring bufferBuffer RED but no escalation"Buffer is RED — recovery Actions needed now."
Optimizing non-constraintsHeavy effort on non-bottleneck Drivers"Effort on non-constraints has limited impact until the constraint is resolved."
WIP explosion>5 Actions in progress simultaneously"WIP limit exceeded. Complete or pause Actions before starting new ones."
Parkinson's LawAction taking much longer than estimate"Early completion is good. Hand off immediately — don't expand work to fill time."

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.