generate-agentic-loop
/generate-agentic-loop
Design or implement bounded, self-correcting agent execution loops with verification, resource limits, and recoverable state. Use when a user asks to build or improve an agent loop or its execution harness.
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 29.10.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
What it does and when to use it
Design or implement bounded, self-correcting agent execution loops with verification, resource limits, and recoverable state. Use when a user asks to build or improve an agent loop or its execution harness.
Original author description
Design or implement bounded, self-correcting agent execution loops with verification, resource limits, and recoverable state. Use when a user asks to build or improve an agent loop or its execution harness.
How it works
SKILL.md has no level-two headings · 16 lines, 398 words
SKILL.md preview
--- name: generate-agentic-loop description: Design or implement bounded, self-correcting agent execution loops with verification, resource limits, and recoverable state. Use when a user asks to build or improve an agent loop or its execution harness. --- # Generate agentic loops Translate the user's objective into an executable loop and an operational harness. Adapt to the requested language, platform, tools, and risk; do not assume Python or background processes are available. 1. Define the objective, scope, authorized actions, measurable acceptance criteria, and independent verification. Identify dependencies and assumptions. If success cannot be objectively verified, report the limitation and require a suitable review instead of claiming success. 2. Select a pattern appropriate to the task: maker-checker, tool/error feedback, or a long-running loop with persisted state. Separate the agent's observe → act → verify → refine cycle from the harness that enforces permissions, limits, and tool execution. 3. Define explicit terminal outcomes: success, iteration limit, budget limit, stall, blocker, interruption, and execution error. Success requires all acceptance checks to pass; a model's unsupported PASS is not sufficient. Return the outcome, evidence, last verified result, and remaining work. 4. Set finite iteration and time limits and, when usage can be measured, token and cost limits. Check limits before each operation and account for actual usage afterward; document any overshoot or unenforceable limit. Bound tool calls with timeouts. Never claim a hard cost cap when the provider cannot enforce one. 5. Preserve actionable failure evidence, including exact errors where safe, and use it to choose the next action. Detect repeated failures and unchanged outputs, but do not mistake an identical output for a stall before evaluating it. Halt after a configurable consecutive repeat threshold; distinguish transient failures from missing permissions or credentials. Do not retry irreversible actions without idempotency or reconciliation. 6. Compact context into an objective, constraints, verified facts and evidence references, current failures, recent attempts, and next action. Keep essential instructions and authorization boundaries. Store large artifacts externally only where available and authorized; redact secrets and apply appropriate retention. Summaries must not invent verified facts. Background memory curation is optional and must not silently change governing instructions. 7. Implement and test the requested solution. Cover successful verification, failed verification, exhaustion, repeated failure, interruption, and budget accounting as applicable. State what was tested and what remains unverified. Never represent a template as production-ready without integration and operational validation. When producing a design, include the chosen architecture, state and action interfaces, verification evidence, stop conditions, resource enforcement, recovery behavior, and assumptions. When implementing, adapt these contracts to the actual environment and expose configuration rather than inventing credentials or permissions.
Versions
The database has no other skills named generate-agentic-loop.
Database snapshot: 13.11.2026 · package version ef068f8999f6 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100loop-engineering
Implements self-correcting agent execution loops, multi-agent parallel workflows, workspace isolation (Git Worktrees), and automated cron-based Pull Request auditing. Use this skill when asked to orchestrate parallel workers or construct review loops. - 100loop-engineering
Implements self-correcting agent execution loops, multi-agent parallel workflows, workspace isolation (Git Worktrees), and automated cron-based Pull Request auditing. Use this skill when asked to orchestrate parallel workers or construct review loops. - 100harness-engineering
Design, build, diagnose, and evolve agent harnesses: instructions, tools, execution environments, durable state, context management, verification, recovery, and bounded autonomous loops. Use for unreliable coding agents, cross-session drift, premature completion, harness audits,… - 100agentic-workflow-builder
Activates AgenticWorkflowBuilder for designing end-to-end agentic systems and AI-powered automation workflows. Use when you need to design a multi-step agentic pipeline, architect a human-in-the-loop approval system, build an autonomous task execution loop with error recovery,… - 100loop-library
Discover, find, compare, audit, repair, adapt, and design repeatable AI-agent loops with explicit triggers, actions, verification, stopping conditions, guardrails, and handoffs. Use when a user asks to analyze a codebase for potential loops, mine coding-thread history for work… - 90self-improving-agent
Patterns for AI agents that learn from their own execution, detect failure modes, and improve autonomously. Use when building agents that get better over time, managing auto- memory, or designing self-correcting feedback loops.