{
  "schemaVersion": "1",
  "revision": "22ce5d34dbcb5a78",
  "id": "factory-evolving-principles-practice",
  "slug": "evolving-principles-practice",
  "title": "Which principles survive better agents? — prompts and exercises",
  "summary": "Complete reusable prompts, setup recipes, success checks, and caveats for which principles survive better agents?.",
  "category": "Principles",
  "tags": [
    "prompts",
    "exercises",
    "evolving-principles"
  ],
  "sources": [
    {
      "label": "Which principles survive better agents?",
      "url": "https://hallowed-entry-96xv.here.now/"
    },
    {
      "label": "Public setup contract",
      "url": "https://hallowed-entry-96xv.here.now/agent.md"
    }
  ],
  "sourceUnits": [
    "00-evolving/index.html#Quick start",
    "00-evolving/index.html#Prompt library",
    "00-evolving/index.html#Set it up (agent guide)",
    "00-evolving/index.html#Run the 14-directive self-audit",
    "00-evolving/index.html#Sort your habits: judgment or workaround",
    "00-evolving/index.html#Fact-check any claim against your own history",
    "00-evolving/agent.md"
  ],
  "collection": "factory",
  "categoryId": "principles",
  "wordCount": 2501,
  "url": "/topics/factory-evolving-principles-practice/",
  "markdownUrl": "/markdown/factory-evolving-principles-practice.md?raw=1&v=22ce5d34dbcb5a78",
  "jsonUrl": "/api/v1/documents/factory-evolving-principles-practice.json",
  "headings": [
    {
      "id": "quick-start",
      "text": "Quick start",
      "level": 2
    },
    {
      "id": "prompt-library",
      "text": "Prompt library",
      "level": 2
    },
    {
      "id": "run-the-14-directive-self-audit",
      "text": "Run the 14-directive self-audit",
      "level": 3
    },
    {
      "id": "sort-your-habits-judgment-or-workaround",
      "text": "Sort your habits: judgment or workaround",
      "level": 3
    },
    {
      "id": "fact-check-any-claim-against-your-own-history",
      "text": "Fact-check any claim against your own history",
      "level": 3
    },
    {
      "id": "set-it-up-agent-guide",
      "text": "Set it up (agent guide)",
      "level": 2
    },
    {
      "id": "complete-reusable-setup-contract",
      "text": "Complete reusable setup contract",
      "level": 2
    },
    {
      "id": "how-agentic-work-is-evolving-agent-setup-guide",
      "text": "How Agentic Work Is Evolving — agent setup guide",
      "level": 2
    },
    {
      "id": "prerequisites",
      "text": "Prerequisites",
      "level": 2
    },
    {
      "id": "steps",
      "text": "Steps",
      "level": 2
    },
    {
      "id": "success-check",
      "text": "Success check",
      "level": 2
    },
    {
      "id": "notes-for-agents",
      "text": "Notes for agents",
      "level": 2
    }
  ],
  "markdown": "# Which principles survive better agents? — prompts and exercises\n\nComplete reusable prompts, setup recipes, success checks, and caveats for which principles survive better agents?.\n\n## Quick start\n\nSize up your own evidence before the audit. Swap paths for wherever your history actually lives.\n\n``` text\n# search your past agent conversations (cs = any full-text index over your chats)\ncs compaction --recent -n 20\n# newest prototypes, and the depth of the well\nls -dt ~/dev/*/ | head -5\nls -d ~/dev/*/ | wc -l\n# archived agent sessions available to mine\nls ~/.oracle/sessions | wc -l\n```\n\n## Prompt library\n\n### Run the 14-directive self-audit\n\n- Agent: any coding agent with shell access\n- Working directory: anywhere on your machine\n- Expected result: a 14-row scorecard table with dated evidence and verdicts\n\n``` text\nRun the 14-directive self-audit of my agentic workflow against my own machine.\n\nBackground: \"The Field Guide — Coding with Agents\" (https://kindly-mirage-ejqd.here.now/) names 14 Prime Directives and calls them \"the layer that doesn't rot.\" Audit that claim against MY machine: for each directive, find dated local evidence and give a verdict.\n\nEvidence sources (adapt to what actually exists here):\n1. My agent conversation history. Use a full-text index if I have one (e.g. `cs <terms> --recent -n 20`); otherwise grep the raw session archives (Claude Code: ~/.claude/projects/, Codex: ~/.codex/sessions/, plus any others you find).\n2. My prototype directories: `ls -dt ~/dev/*/ | head -30` (or wherever my projects live). Count them; note naming patterns like repeated \"-final\" suffixes.\n3. Any other session archives at scale — count them so we know the size of the well.\n\nThe 14 directives, with what to look for:\n01 \"If the agent can't see it, it doesn't exist.\" — do my agents find their own context, or do I hand-feed it?\n02 \"Research is loud. Contracts are quiet.\" — do I plan separately from executing? What enforces alignment?\n03 \"If you can pipe it, you can reuse it.\" — is my piping genuine composition, or leftover context management?\n04 \"Commit before autonomy.\" — is Git carrying my prototypes, or do directory names and published sites version things?\n05 \"Done is a claim. Green is a fact.\" — do finished tasks leave saved checks the next agent can run?\n06 \"The agent can't read your mind — only your artifacts.\" — how does intent enter: typing, dictation, files?\n07 \"Expose the decision space. Keep the decision.\" — do alternatives arrive as prose, code, or generated images?\n08 \"Bet small. Learn fast. Revert free.\" — how big is the prototype discard pile, and do I ever mine it?\n09 \"Reuse follows value — never precedes it.\" — how many capture tools do I run (instruction files, skills, memory dirs), and which have gone stale?\n10 \"No single switch makes autonomy safe.\" — what actually guards unsupervised runs: allowlists, backups, a secrets manager?\n11 \"You can launch more agents than you can review.\" — how many parallel agent tracks do I actually sustain?\n12 \"Boring code is a superpower.\" — what is the most novel tool stack in my recent sessions?\n13 \"A gate beats a plea.\" — who wrote my hooks and permission config: me, or an agent?\n14 \"The code is disposable. The contract is not.\" — where do my sessions show selection and taste as the real bottleneck?\n\nOutput: one table row per directive — # | law (quoted) | evidence from my machine (dated: session, file, or directory) | verdict. Verdict vocabulary: Held / Hardened / Partly inverted / Outgrew tooling.\n\nRules: every verdict needs at least one dated, citable artifact from this machine — no verdicts from vibes. If a source doesn't exist here, write that in the row instead of inventing evidence. Never include tokens, emails, or private identifiers in the output. Close with a 3-line summary: counts per verdict, and which of my habits are tool-shaped workarounds due for retirement.\n```\n\n### Sort your habits: judgment or workaround\n\n- Agent: any coding agent with shell access\n- Working directory: your home directory or main project\n- Expected result: 10 habits, each bucketed, with keep / retool / retire recommendations\n\n``` text\nSearch my agent conversation history and my project instruction files (CLAUDE.md, AGENTS.md, hooks, saved prompts, skills) and list my 10 most ingrained agent-workflow habits — rules I repeat, ceremonies I enforce, rigs I keep rebuilding.\n\nSort each habit into exactly one bucket:\nA. Encodes human judgment (verification, review limits, taste) — keep it; consider promoting it to a deterministic gate.\nB. Compensates for a tool weakness — name the weakness explicitly (e.g. \"context windows were small\", \"compaction lost my work\").\n\nFor every bucket-B habit, check whether the named weakness still exists in my current tools, and find the last dated time it actually bit me. Recommend keep, retool, or retire for each. A habit I can't justify without naming a earlier-generation tool limit is due for retirement.\n```\n\n### Fact-check any claim against your own history\n\n- Agent: any coding agent with shell access\n- Working directory: anywhere on your machine\n- Expected result: a verdict with dated citations from your own sessions and files\n\n``` text\nFact-check this claim against my own machine: \"<paste any claim about how you work, e.g. 'I always commit before letting an agent run unsupervised'>\".\n\nSearch my agent conversation history for dated sessions that support or contradict it, scan my project directories for matching artifacts, and return a verdict with citations (dates plus file, session, or directory names). If the evidence is thinner than the claim, say so — an agent saying \"confirmed\" is not evidence; the citations are.\n```\n\n## Set it up (agent guide)\n\n1.  Inventory your evidence sources. Run these and note what exists:\n\n    ``` text\n    command -v cs || echo \"no conversation index — will grep raw archives\"\n    ls -d ~/.claude/projects ~/.codex/sessions ~/.oracle/sessions 2>/dev/null\n    ls -d ~/dev 2>/dev/null || echo \"adjust to your projects directory\"\n    ```\n\n2.  Create the scorecard skeleton in your working directory:\n\n    ``` text\n    # 14-Directive Self-Audit — <date>\n\n    Audited machine: <describe generically — never record private paths, tokens, or ids>\n    Sources searched: <conversation index / session archives / prototype directory>\n\n    | #  | Old law (quoted from the Field Guide) | Evidence from this machine (dated) | Verdict |\n    |----|----|----|----|\n    | 01 | \"If the agent can't see it, it doesn't exist.\" | | |\n    | 02 | \"Research is loud. Contracts are quiet.\" | | |\n    | 03 | \"If you can pipe it, you can reuse it.\" | | |\n    | 04 | \"Commit before autonomy.\" | | |\n    | 05 | \"Done is a claim. Green is a fact.\" | | |\n    | 06 | \"The agent can't read your mind — only your artifacts.\" | | |\n    | 07 | \"Expose the decision space. Keep the decision.\" | | |\n    | 08 | \"Bet small. Learn fast. Revert free.\" | | |\n    | 09 | \"Reuse follows value — never precedes it.\" | | |\n    | 10 | \"No single switch makes autonomy safe.\" | | |\n    | 11 | \"You can launch more agents than you can review.\" | | |\n    | 12 | \"Boring code is a superpower.\" | | |\n    | 13 | \"A gate beats a plea.\" | | |\n    | 14 | \"The code is disposable. The contract is not.\" | | |\n\n    ## Summary\n\n    - Held or hardened: __ / 14\n    - Partly inverted: __ / 14\n    - Outgrew tooling: __ / 14\n    - Tool-shaped habits due for retirement:\n      -\n    ```\n\n3.  Fill every row: for each directive, search your history and record at least one dated artifact plus a verdict (Held / Hardened / Partly inverted / Outgrew tooling). With an index, use it; without one, grep the raw archives:\n\n    ``` text\n    cs \"compaction\" --recent -n 20\n    grep -rl \"compaction\" ~/.claude/projects --include='*.jsonl' 2>/dev/null | head -5\n    ```\n\n**Success check:** the scorecard has exactly 14 directive rows, every Verdict cell is filled, and every Evidence cell contains a date. The count only includes rows with a non-empty Verdict cell, so an unfilled scorecard fails.\n\n``` text\ngrep -cE '^\\| (0[1-9]|1[0-4]) \\|([^|]*\\|){2} *[^ |]' agent-audit/scorecard.md\n```\n\n``` text\n14\n```\n\n## Complete reusable setup contract\n\n## How Agentic Work Is Evolving — agent setup guide\n\nFollowing this file sets up: a 14-directive self-audit of your user's agentic workflow, run against their own machine, producing a filled-in scorecard file (`agent-audit/scorecard.md`).\n\nBackground: \"The Field Guide — Coding with Agents\" (https://kindly-mirage-ejqd.here.now/) names 14 Prime Directives and calls them \"the layer that doesn't rot.\" Your job is to audit that claim against THIS machine: for each directive, find dated local evidence and assign a verdict.\n\nPlaceholders used below: `<project-root>` = the directory you are working in. All other paths are literal but may not exist on every machine — adapt per the Notes for agents.\n\nHuman presentation pages for this section (context for your user, not required for the audit): the section guide at `index.html`, a Session Replay of the audit session at `replay.html`, and a Socratic Dialogue at `dialogue.html` — all at this site's root. This file remains the canonical machine copy of the setup steps.\n\n## Prerequisites\n\nCheck each before starting:\n\n```bash\ncommand -v cs || echo \"no conversation index — will grep raw archives\"\nls -d ~/.claude/projects ~/.codex/sessions ~/.oracle/sessions 2>/dev/null\nls -d ~/dev 2>/dev/null || echo \"adjust to your user's projects directory\"\n```\n\nExpected: at least ONE source of agent history (a conversation index like `cs`, or a raw session archive directory) and one projects directory. If none exist, stop and tell your user there is no local history to audit — do not invent evidence.\n\n## Steps\n\n1. Create `<project-root>/agent-audit/scorecard.md` with exactly this content:\n\n   ```markdown\n   # 14-Directive Self-Audit — <date>\n\n   Audited machine: <describe generically — never record private paths, tokens, or ids>\n   Sources searched: <conversation index / session archives / prototype directory>\n\n   | #  | Old law (quoted from the Field Guide) | Evidence from this machine (dated) | Verdict |\n   |----|----|----|----|\n   | 01 | \"If the agent can't see it, it doesn't exist.\" | | |\n   | 02 | \"Research is loud. Contracts are quiet.\" | | |\n   | 03 | \"If you can pipe it, you can reuse it.\" | | |\n   | 04 | \"Commit before autonomy.\" | | |\n   | 05 | \"Done is a claim. Green is a fact.\" | | |\n   | 06 | \"The agent can't read your mind — only your artifacts.\" | | |\n   | 07 | \"Expose the decision space. Keep the decision.\" | | |\n   | 08 | \"Bet small. Learn fast. Revert free.\" | | |\n   | 09 | \"Reuse follows value — never precedes it.\" | | |\n   | 10 | \"No single switch makes autonomy safe.\" | | |\n   | 11 | \"You can launch more agents than you can review.\" | | |\n   | 12 | \"Boring code is a superpower.\" | | |\n   | 13 | \"A gate beats a plea.\" | | |\n   | 14 | \"The code is disposable. The contract is not.\" | | |\n\n   ## Summary\n\n   - Held or hardened: __ / 14\n   - Partly inverted: __ / 14\n   - Outgrew tooling: __ / 14\n   - Tool-shaped habits due for retirement:\n     -\n   ```\n\n2. Size up the evidence well and record the counts in the scorecard header area:\n\n   ```bash\n   ls -d ~/dev/*/ 2>/dev/null | wc -l\n   ls ~/.oracle/sessions 2>/dev/null | wc -l\n   ls -dt ~/dev/*/ 2>/dev/null | head -10\n   ```\n\n3. Gather evidence for each of the 14 directives. What to look for, per row:\n\n   - 01 — do the user's agents find their own context, or does the user hand-feed it? Search history for context-curation rituals vs agent-driven search.\n   - 02 — is planning separated from execution? What (if anything) enforces alignment: task files, checkpoint agents, goal contracts?\n   - 03 — is piping between agents genuine composition (CI, scripts), or leftover context management from when compaction was bad?\n   - 04 — is Git carrying the prototypes, or do directory names (`*-final`, `*-super-final`) and published sites act as versioning?\n   - 05 — do finished tasks leave saved checks the next agent can run?\n   - 06 — how does intent enter: typing, dictation, artifacts? Search for dictation/brain-dump patterns.\n   - 07 — do alternatives arrive as prose, code variants, or generated images?\n   - 08 — how big is the prototype discard pile, and is it ever mined for ideas?\n   - 09 — how many success-capture tools are in use (instruction files, skills, memory dirs), and which have gone stale?\n   - 10 — what guards unsupervised runs: allowlists, backups, a secrets manager like the `op` CLI?\n   - 11 — how many parallel agent tracks does the user actually sustain?\n   - 12 — what is the most novel tool stack in recent sessions?\n   - 13 — who wrote the hooks and permission config: the user, or an agent?\n   - 14 — where do sessions show selection and taste as the real bottleneck?\n\n   Search with the index if one exists, otherwise grep the raw archives:\n\n   ```bash\n   cs \"compaction\" --recent -n 20\n   grep -rl \"compaction\" ~/.claude/projects --include='*.jsonl' 2>/dev/null | head -5\n   ```\n\n4. Fill every row: at least one dated, citable artifact (session, file, or directory) in the Evidence column, and one verdict from the fixed vocabulary — `Held`, `Hardened`, `Partly inverted`, `Outgrew tooling`. Qualifiers are fine (e.g. \"Held — mechanism moved\"). Then fill in the Summary counts and list any tool-shaped habits due for retirement (habits that only made sense given a earlier-generation tool weakness that no longer exists).\n\n## Success check\n\nRun:\n\n```bash\ngrep -cE '^\\| (0[1-9]|1[0-4]) \\|([^|]*\\|){2} *[^ |]' agent-audit/scorecard.md\n```\n\nExpected observable result: `14`. Only directive rows with a non-empty Verdict cell count, so an unfilled or half-filled scorecard fails this check. Also confirm by inspection that every Evidence cell contains a date (or an explicit \"no local evidence found\" note).\n\nReport to your user: the finished scorecard table, the summary counts, which habits look due for retirement, and anything you had to adapt (missing sources, substituted paths).\n\n## Notes for agents\n\n- `cs` is a personal full-text conversation index on the original author's machine. If it is missing, grep the raw session archives instead (Claude Code stores JSONL under `~/.claude/projects/`; other CLIs have their own directories). Any searchable history works.\n- `~/dev` and `~/.oracle/sessions` are the original author's layout. Substitute wherever your user keeps projects and session archives; ask the user if you cannot find them.\n- Evidence discipline: a verdict without a dated, citable artifact is a vibe, not a verdict. Write \"no local evidence found\" in the row rather than inventing something.\n- Privacy: the scorecard may be shared later. Never write tokens, emails, session IDs, or private absolute paths into it — describe artifacts generically (\"a the recorded workflow session\", \"a project directory named `<name>-final`\").\n- If the machine has very little history (new machine, fresh install), report that the audit is not meaningful here instead of forcing 14 verdicts."
}
