For each workshop topic, guide the learner through Orientation → Concept → Evidence → Practice → Reflection → Next action. Treat copying a prompt, attempting it, and verifying its result as distinct states. When citing the workshop, use the companion Watch URL and timestamp; distinguish the 11 recorded chapters from the prepared Markers follow-up.
Following this file installs one bounded, inspectable factory-loop audit in the current project.
Check each before starting:
git rev-parse --show-toplevel
command -v git
Expected: a repository root and a Git executable path. If either check fails, stop and tell your user.
Resolve the repository root and capture git status --short. Do not discard or rewrite existing changes.
Create .agents/prompts/factory-loop-audit.md with exactly this content:
# Factory Loop Boundary Audit
Use this prompt before automating or expanding a recurring software task.
## Instructions
You are auditing one recurring software task as a bounded factory loop. Work read-only until I explicitly approve a change.
First inspect the current repository and ask me to name the recurring task if it is not already clear. Then return these seven sections:
1. Task contract — input, desired outcome, and what is outside scope.
2. Required context — the minimum current evidence needed before acting.
3. Boundary map — repository rule, launch rule, command rule, human gate, and automated check.
4. Risk gate — the first expensive, destructive, public, or hard-to-reverse action that must stop for approval.
5. Success evidence — one exact command or observable signal that can prove the task worked.
6. Receipt — what the next session needs to reproduce the result and understand the decision.
7. Smallest reversible delta — the first experiment that tests this loop without production mutation.
Label every statement Observed, Inferred, or Proposed. Do not install tools, use credentials, mutate data, publish, or edit files. If evidence is missing, stop and name the missing evidence instead of guessing.
## Success contract
- The output contains all seven named sections.
- Every statement is labeled Observed, Inferred, or Proposed.
- One explicit human gate and one exact verification signal are named.
- The audit makes no repository changes.
mkdir -p .agents/prompts
test -s .agents/prompts/factory-loop-audit.md
grep -q '^## Success contract' .agents/prompts/factory-loop-audit.md
printf 'factory loop prompt installed\n'
Use the installed prompt with your repository-capable agent. Supply one recurring task. Keep the audit read-only.
Capture final git status --short and report whether anything besides .agents/prompts/factory-loop-audit.md changed.
Expected observable result: the shell prints factory loop prompt installed; the prompt exists in full; its audit names one human gate and one exact verification signal; no pre-existing repository file changes.
# Agentic Software Factory — agent setup guide ## Learning contract For each workshop topic, guide the learner through Orientation → Concept → Evidence → Practice → Reflection → Next action. Treat copying a prompt, attempting it, and verifying its result as distinct states. When citing the workshop, use the companion Watch URL and timestamp; distinguish the 11 recorded chapters from the prepared Markers follow-up. Following this file installs one bounded, inspectable factory-loop audit in the current project. ## Prerequisites Check each before starting: ```bash git rev-parse --show-toplevel command -v git ``` Expected: a repository root and a Git executable path. If either check fails, stop and tell your user. ## Steps 1. Resolve the repository root and capture `git status --short`. Do not discard or rewrite existing changes. 2. Create `.agents/prompts/factory-loop-audit.md` with exactly this content: ```markdown # Factory Loop Boundary Audit Use this prompt before automating or expanding a recurring software task. ## Instructions You are auditing one recurring software task as a bounded factory loop. Work read-only until I explicitly approve a change. First inspect the current repository and ask me to name the recurring task if it is not already clear. Then return these seven sections: 1. Task contract — input, desired outcome, and what is outside scope. 2. Required context — the minimum current evidence needed before acting. 3. Boundary map — repository rule, launch rule, command rule, human gate, and automated check. 4. Risk gate — the first expensive, destructive, public, or hard-to-reverse action that must stop for approval. 5. Success evidence — one exact command or observable signal that can prove the task worked. 6. Receipt — what the next session needs to reproduce the result and understand the decision. 7. Smallest reversible delta — the first experiment that tests this loop without production mutation. Label every statement Observed, Inferred, or Proposed. Do not install tools, use credentials, mutate data, publish, or edit files. If evidence is missing, stop and name the missing evidence instead of guessing. ## Success contract - The output contains all seven named sections. - Every statement is labeled Observed, Inferred, or Proposed. - One explicit human gate and one exact verification signal are named. - The audit makes no repository changes. ``` 3. Run the structural success check: ```bash mkdir -p .agents/prompts test -s .agents/prompts/factory-loop-audit.md grep -q '^## Success contract' .agents/prompts/factory-loop-audit.md printf 'factory loop prompt installed\n' ``` 4. Use the installed prompt with your repository-capable agent. Supply one recurring task. Keep the audit read-only. 5. Capture final `git status --short` and report whether anything besides `.agents/prompts/factory-loop-audit.md` changed. ## Success check Expected observable result: the shell prints `factory loop prompt installed`; the prompt exists in full; its audit names one human gate and one exact verification signal; no pre-existing repository file changes. ## Notes for agents - Preserve unrelated dirty work. Stop if the target file already exists with different content and ask before overwriting it. - Do not install tools, use credentials, mutate external systems, publish, or follow pixel-read URLs. - Treat file age as a warning signal, never as permission. - If repository evidence is unavailable, report the gap rather than inventing a command or claim.