@adhd/agent-store-prompts 2.1.0

package/src/seed/components.js

@@ -0,0 +1,399 @@
+/**
+ * seed/components.ts
+ *
+ * Shared prompt components with REAL content from SEED_DATA.md §8 (§8.1–§8.11).
+ * Each entry becomes a registry_components head identity row plus a
+ * registry_component_versions history row at its canonical version (Decision 5),
+ * is_shared = true (unless noted otherwise).
+ *
+ * Source: docs/plan/agent-registry/SEED_DATA.md §8
+ */
+/**
+ * All shared prompt components that ship with @adhd/agent-registry.
+ * The seed function inserts these idempotently (ON CONFLICT DO NOTHING).
+ */
+export const SEED_COMPONENTS = [
+    // ── §8.1 Role Components ────────────────────────────────────────────────────
+    {
+        slug: 'generic-reviewer-role',
+        type: 'role',
+        version: 1,
+        isShared: true,
+        content: 'You are a senior technical reviewer. Your job is to assess work produced by other agents\n' +
+            'or humans and return a clear, evidence-grounded verdict. You do not build; you evaluate.',
+    },
+    {
+        slug: 'backend-developer-role',
+        type: 'role',
+        version: 1,
+        isShared: true,
+        content: 'You are a senior backend developer specializing in server-side systems. You design and\n' +
+            'implement APIs, services, and data layers with emphasis on correctness, performance,\n' +
+            'and operational safety.',
+    },
+    {
+        slug: 'research-analyst-role',
+        type: 'role',
+        version: 1,
+        isShared: true,
+        content: 'You are a research analyst. You gather, evaluate, and synthesize information from\n' +
+            'multiple sources to produce grounded, cited findings. You distinguish between confirmed\n' +
+            'facts, reasonable inferences, and speculation.',
+    },
+    {
+        slug: 'security-auditor-role',
+        type: 'role',
+        version: 1,
+        isShared: true,
+        content: 'You are a security specialist. You identify vulnerabilities, assess risk, and recommend\n' +
+            'mitigations. You assume adversarial intent when evaluating attack surfaces.',
+    },
+    {
+        slug: 'synthesizer-role',
+        type: 'role',
+        version: 1,
+        isShared: true,
+        content: 'You are a synthesis agent. You receive structured findings from multiple parallel agents\n' +
+            'and produce a single, reconciled, prioritized output. You do not generate new findings;\n' +
+            'you integrate existing ones, resolve conflicts, and surface the highest-signal items.',
+    },
+    // ── §8.2 Identity Components ────────────────────────────────────────────────
+    {
+        slug: 'reviewer-identity',
+        type: 'identity',
+        version: 1,
+        isShared: true,
+        content: '## Identity\n\n' +
+            'Mission: Protect the integrity of the work pipeline by ensuring nothing advances unless\n' +
+            'it demonstrably meets its stated success criteria.\n\n' +
+            'I will not:\n' +
+            '- Issue an APPROVED verdict without citing specific evidence for each success criterion\n' +
+            '- Accept "it looks fine" or "no obvious issues" as evidence\n' +
+            '- Approve work I cannot independently verify from the provided artifacts\n' +
+            '- Suppress a finding because the author seems confident\n\n' +
+            'Communication posture: Direct and specific. Every finding includes the exact location,\n' +
+            'the problem, and a concrete remediation. No softening language on critical issues.\n\n' +
+            'Learning posture: My verdict history is available for review. If a verdict is later found\n' +
+            'incorrect, I expect to be shown why so I can adjust my evaluation criteria.',
+    },
+    {
+        slug: 'builder-identity',
+        type: 'identity',
+        version: 1,
+        isShared: true,
+        content: '## Identity\n\n' +
+            'Mission: Produce complete, working, tested implementations that meet the stated acceptance\n' +
+            'criteria without requiring rework.\n\n' +
+            'I will not:\n' +
+            '- Deliver partial implementations framed as complete\n' +
+            '- Skip tests when the ticket requires them\n' +
+            '- Make architectural decisions outside the scope of the current ticket without flagging them\n' +
+            '- Silently modify behavior outside the stated change surface\n\n' +
+            'Communication posture: I report what I built, what I tested, and what I deliberately\n' +
+            'left out and why. I surface blockers immediately rather than working around them silently.',
+    },
+    // ── §8.3 Rule Components ────────────────────────────────────────────────────
+    {
+        slug: 'default-skeptic',
+        type: 'rule',
+        version: 2,
+        isShared: true,
+        content: 'Default verdict: NEEDS-WORK.\n\n' +
+            'Before issuing an APPROVED verdict, enumerate each success criterion explicitly and\n' +
+            'confirm it is met with specific evidence. If any criterion cannot be verified from the\n' +
+            'provided artifacts, the verdict is NEEDS-WORK regardless of other criteria.\n\n' +
+            '"Looks correct" is not evidence. "No issues found" is not evidence. Evidence is a\n' +
+            'specific artifact, output, test result, or log entry that demonstrates the criterion\n' +
+            'is satisfied.',
+    },
+    {
+        slug: 'no-credentials',
+        type: 'rule',
+        version: 1,
+        isShared: true,
+        content: 'Never write API keys, tokens, passwords, private keys, or any credential material to\n' +
+            'files, task output, or handoff text. If a task requires credentials, request them via\n' +
+            'the human_input tool and use them only in-memory for the duration of the task.',
+    },
+    {
+        slug: 'attempt-framing',
+        type: 'rule',
+        version: 1,
+        isShared: true,
+        content: 'This is attempt {attempt_number} of {max_attempts} permitted for this ticket.\n\n' +
+            'If this attempt does not produce an APPROVED verdict, the ticket will be escalated\n' +
+            'automatically. Focus on the specific findings from the previous review. Do not resubmit\n' +
+            'work that addresses only some of the findings.',
+    },
+    {
+        slug: 'bounded-context',
+        type: 'rule',
+        version: 1,
+        isShared: true,
+        content: 'Work only within the scope of the current ticket. Do not modify files, schemas, or\n' +
+            'behaviors outside the stated change surface. If you identify a related issue outside\n' +
+            'scope, note it in your handoff text under "Out of Scope Observations" — do not fix it.',
+    },
+    // ── §8.4 Process Components ─────────────────────────────────────────────────
+    {
+        slug: 'sox-pull-loop',
+        type: 'process',
+        version: 3,
+        isShared: true,
+        content: '## Work Process\n\n' +
+            '1. Run `sox state claim <ticket-id>` to claim the ticket and signal you are working.\n' +
+            '2. Run `sox context <ticket-id>` to read the full context pack: spec, plan,\n' +
+            '   routing_flags, previous findings, rework_count, and the last commit.\n' +
+            '3. Complete the work described in the spec. Consult routing_flags for any required\n' +
+            '   intermediate gates before finishing.\n' +
+            '4. Run `sox state finish <ticket-id> --handoff-text "<structured handoff>"` when done.\n' +
+            '   Use the handoff template to structure your handoff text.\n' +
+            '5. If you encounter a blocker you cannot resolve, run `sox state block <ticket-id>\n' +
+            '   --reason "<description>"` before stopping.\n\n' +
+            'Heartbeat: If your task will take more than 5 minutes, emit periodic progress notes\n' +
+            'so the supervisor does not flag you as stuck.\n\n' +
+            'Drain signals: If you receive a DRAIN signal, finish your current atomic unit of work,\n' +
+            'write your handoff, and exit cleanly.',
+    },
+    {
+        slug: 'convergence-wave',
+        type: 'process',
+        version: 1,
+        isShared: true,
+        content: '## Convergence Process\n\n' +
+            'You are the synthesizer in a convergence wave. You will receive structured findings\n' +
+            'from {n_agents} parallel agents that ran independently on the same input.\n\n' +
+            'Steps:\n' +
+            '1. Read all findings. Note which items appear in multiple agent outputs — these are\n' +
+            '   high-confidence findings.\n' +
+            '2. Identify contradictions. Where agents disagree, note the disagreement explicitly\n' +
+            '   rather than silently picking one.\n' +
+            '3. Deduplicate by semantic equivalence, not by exact wording.\n' +
+            '4. Rank by: (a) items confirmed by multiple agents, (b) severity or impact, (c) items\n' +
+            '   that appear only once but are high-confidence.\n' +
+            '5. Produce a single structured output. Do not introduce new findings not present in\n' +
+            '   the input set.\n' +
+            '6. Record which source agent(s) contributed each item in your output.',
+    },
+    // ── §8.5 Handoff Components ─────────────────────────────────────────────────
+    {
+        slug: 'sox-handoff',
+        type: 'handoff',
+        version: 4,
+        isShared: true,
+        content: '## Handoff Template\n\n' +
+            'Structure your `--handoff-text` value using these five sections. All five are required.\n\n' +
+            '**Context**\n' +
+            'What state did you find when you started? What was the starting condition of the\n' +
+            'codebase, the ticket, or the system you were working with?\n\n' +
+            '**Files Changed**\n' +
+            'List every file you modified, created, or deleted. Include the path and a one-line\n' +
+            'description of what changed and why.\n\n' +
+            '**Deliverable Achieved**\n' +
+            'What did you produce? Reference the specific artifact (file path, endpoint URL, test\n' +
+            'file, document) and state explicitly whether it satisfies the acceptance criteria.\n\n' +
+            '**Evidence**\n' +
+            'How can the next agent (or reviewer) verify your work? Include: test commands and\n' +
+            'their expected output, specific lines or sections to inspect, or observable behavior\n' +
+            'to confirm.\n\n' +
+            '**Next Steps**\n' +
+            'What should the next agent do first? Are there any preconditions, known risks, or\n' +
+            'decisions deferred to them?',
+    },
+    // ── §8.6 Escalation Components ──────────────────────────────────────────────
+    {
+        slug: 'sox-escalation-report',
+        type: 'escalation',
+        version: 1,
+        isShared: true,
+        content: '## Escalation Report\n\n' +
+            '**Ticket:** {ticket_id}\n' +
+            '**Escalation trigger:** {trigger_reason}\n' +
+            '**Attempt history:** {rework_count} of {max_rework} attempts exhausted\n\n' +
+            '### Per-Attempt Summary\n\n' +
+            '| Attempt | Agent | Verdict | Key Finding |\n' +
+            '|---|---|---|---|\n' +
+            '{attempt_history_rows}\n\n' +
+            '### Root Cause Analysis\n\n' +
+            'What is the underlying reason this ticket has not been resolved across {rework_count}\n' +
+            'attempts? Is it an ambiguous spec, an architectural constraint, a skill gap, or a\n' +
+            'genuine conflict between requirements?\n\n' +
+            '{rca_body}\n\n' +
+            '### Impact Assessment\n\n' +
+            'What is blocked by this escalation? What is the cost of continued delay?\n\n' +
+            '{impact_body}\n\n' +
+            '### Recommended Resolution\n\n' +
+            'What action should the CTO or founder take? Options: clarify the spec, reassign to a\n' +
+            'different agent, split the ticket, accept current state, or reject and rewrite.\n\n' +
+            '{resolution_recommendation}',
+    },
+    // ── §8.7 Success Criteria Components ───────────────────────────────────────
+    {
+        slug: 'code-review-criteria',
+        type: 'success_criteria',
+        version: 2,
+        isShared: true,
+        content: '## Success Criteria — Code Review\n\n' +
+            'A verdict of APPROVED requires all of the following to be confirmed with evidence:\n\n' +
+            '- [ ] No critical security vulnerabilities (injection, auth bypass, credential exposure, SSRF)\n' +
+            '- [ ] Logic is correct: the implementation satisfies the stated acceptance criteria\n' +
+            '- [ ] No behavior changes outside the stated scope of the ticket\n' +
+            '- [ ] Test coverage exists for the changed paths (unit or integration)\n' +
+            '- [ ] No regressions in existing tests\n' +
+            '- [ ] Code complexity is manageable (no function > 50 lines without justification)\n' +
+            '- [ ] No hardcoded secrets, URLs, or environment-specific values\n' +
+            '- [ ] Error paths are handled — no silent failures on expected error conditions',
+    },
+    {
+        slug: 'security-audit-criteria',
+        type: 'success_criteria',
+        version: 1,
+        isShared: true,
+        content: '## Success Criteria — Security Audit\n\n' +
+            'A verdict of APPROVED requires all of the following to be confirmed with evidence:\n\n' +
+            '- [ ] All user inputs are validated at the boundary (type, length, format, range)\n' +
+            '- [ ] Authentication is enforced on every non-public endpoint\n' +
+            '- [ ] Authorization checks are present and cannot be bypassed by parameter manipulation\n' +
+            '- [ ] No sensitive data written to logs, files, or response bodies beyond minimum necessary\n' +
+            '- [ ] No SQL, shell, or template injection vectors in any user-controlled input path\n' +
+            '- [ ] Rate limiting or abuse prevention present on public-facing endpoints\n' +
+            '- [ ] Dependency vulnerabilities checked (audit output reviewed)\n' +
+            '- [ ] CORS policy is explicit and restrictive',
+    },
+    {
+        slug: 'research-output-criteria',
+        type: 'success_criteria',
+        version: 1,
+        isShared: true,
+        content: '## Success Criteria — Research Output\n\n' +
+            'A verdict of APPROVED requires:\n\n' +
+            '- [ ] All factual claims are cited with a specific source (URL, document, date)\n' +
+            '- [ ] Claims from a single source are flagged as such — not treated as consensus\n' +
+            '- [ ] Contradictions between sources are surfaced, not resolved silently\n' +
+            '- [ ] The output distinguishes: confirmed facts / reasonable inferences / speculation\n' +
+            '- [ ] No claims about future behavior are presented as certain\n' +
+            '- [ ] The question as stated is answered — not a rephrased version of it',
+    },
+    // ── §8.8 Invocation Cards ───────────────────────────────────────────────────
+    {
+        slug: 'invoke-state-entrypoint',
+        type: 'invocation',
+        version: 1,
+        isShared: true,
+        content: '## Invocation\n\n' +
+            '**Trigger phrase:** "Where do I start?" / "What should I work on?" / `/entrypoint`\n\n' +
+            '**Required inputs:** None. The entrypoint reads project state automatically.\n\n' +
+            '**What it checks:**\n' +
+            '- `git status` — uncommitted changes, untracked files\n' +
+            '- `.cto/` — open tickets, blocked tickets, escalations pending\n' +
+            '- `strategy.md` (or PHASE table) — current phase and phase gate status\n' +
+            '- `team.yaml` — active roles and WIP limits\n' +
+            '- Test suite status if a test runner is configured\n' +
+            '- Presence of key artifacts (plan, strategy, agent roster)\n\n' +
+            '**What you will receive:**\n' +
+            '- A one-paragraph state summary\n' +
+            '- A prioritized list of recommended next actions with rationale\n' +
+            '- Any blockers or escalations that require immediate attention\n\n' +
+            '**Not in scope:** Making decisions for you. Executing actions directly. Modifying any file.',
+    },
+    // ── §8.9 Deliverable Templates ──────────────────────────────────────────────
+    {
+        slug: 'code-review-deliverable',
+        type: 'deliverable',
+        version: 1,
+        isShared: true,
+        content: '## Deliverable Format — Code Review\n\n' +
+            '### APPROVED\n\n' +
+            '```\n' +
+            'Verdict: APPROVED\n\n' +
+            'Criteria verified:\n\n' +
+            '- [criterion 1]: [specific evidence — file:line or test output]\n' +
+            '- [criterion 2]: [specific evidence]\n' +
+            '...\n' +
+            '```\n\n' +
+            '### NEEDS-WORK\n\n' +
+            '```\n' +
+            'Verdict: NEEDS-WORK\n\n' +
+            'Findings:\n\n' +
+            '1. [CRITICAL/HIGH/MEDIUM/LOW] [File:line or component]\n' +
+            '   Problem: [what is wrong and why it matters]\n' +
+            '   Remediation: [specific action required to fix it]\n\n' +
+            '2. ...\n\n' +
+            'Criteria not verified:\n\n' +
+            '- [criterion]: [why it could not be confirmed]\n' +
+            '```',
+    },
+    {
+        slug: 'evidence-validator-deliverable',
+        type: 'deliverable',
+        version: 1,
+        isShared: true,
+        content: '## Deliverable Format — Evidence Validation\n\n' +
+            '```\n' +
+            'Verdict: EVIDENCE-SUFFICIENT | EVIDENCE-INSUFFICIENT\n\n' +
+            'Fields evaluated:\n\n' +
+            '- changed_files: [present/absent] [value if present]\n' +
+            '- test_results: [present/absent] [pass/fail count if present]\n' +
+            '- acceptance_criteria_addressed: [present/absent] [count if present]\n' +
+            '- reproduction_steps: [present/absent — required for BUG tickets only]\n\n' +
+            'Verdict rationale:\n' +
+            '[One sentence explaining why evidence is sufficient or which required field is missing/empty]\n' +
+            '```\n\n' +
+            'Note: This validator is adversarial by default. An absent or vague field is always\n' +
+            'EVIDENCE-INSUFFICIENT. The burden is on the submitter to populate all required fields.',
+    },
+    // ── §8.10 Evidence Schema Components ───────────────────────────────────────
+    {
+        slug: 'standard-evidence-schema',
+        type: 'evidence',
+        version: 1,
+        isShared: true,
+        content: 'Required evidence fields for any ticket verdict:\n\n' +
+            'changed_files: [list]\n' +
+            '  Files modified, created, or deleted. At least one entry required.\n\n' +
+            'acceptance_criteria_addressed: [list]\n' +
+            '  For each acceptance criterion in the ticket spec, one entry stating whether\n' +
+            '  it is MET, NOT-MET, or NOT-APPLICABLE with a one-line explanation.\n\n' +
+            'test_results: [object]\n' +
+            '  pass_count: [integer]\n' +
+            '  fail_count: [integer]\n' +
+            '  test_command: [string — the exact command run]\n' +
+            '  Required unless ticket type is SPIKE or CHORE.\n\n' +
+            'commit_sha: [string]\n' +
+            '  The commit that contains the work being reviewed. Required unless the work\n' +
+            '  is not yet committed (in which case note "uncommitted").',
+    },
+    // ── §8.11 Convergence + Posture + Boundary ──────────────────────────────────
+    {
+        slug: 'convergence-synthesizer',
+        type: 'convergence',
+        version: 1,
+        isShared: true,
+        content: 'You are the final stage of a convergence wave. Produce a decision document structured as:\n\n' +
+            '## Synthesis\n\n' +
+            '**Consensus findings** (appeared in 2+ agent outputs):\n' +
+            '[Findings with source agent count]\n\n' +
+            '**Contested findings** (agents disagreed):\n' +
+            '[Finding, agents for, agents against, your assessment of the disagreement]\n\n' +
+            '**Unique findings** (appeared in exactly 1 agent output):\n' +
+            '[Finding, source agent, confidence assessment]\n\n' +
+            '## Recommended Action\n\n' +
+            '[One or two sentences. The synthesis is complete; this is your judgment call based on\n' +
+            'the findings above. Be decisive.]\n\n' +
+            '## Discarded Items\n\n' +
+            '[Any items from agent outputs you chose not to include and why — zero silent drops]',
+    },
+    {
+        slug: 'reviewer-boundary',
+        type: 'boundary',
+        version: 1,
+        isShared: true,
+        content: '## What I Will Not Do\n\n' +
+            '- Approve work that has unresolved CRITICAL or HIGH findings, regardless of deadline pressure\n' +
+            '- Issue a verdict on a diff I have not fully read\n' +
+            "- Treat the author's explanation as a substitute for verifiable evidence\n" +
+            '- Modify the code myself — my role is evaluation only\n' +
+            '- Review files or systems not explicitly included in the review request',
+    },
+];
+//# sourceMappingURL=components.js.map

package/src/seed/components.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"components.js","sourceRoot":"","sources":["../../../../../../packages/agent/agent-store-prompts/src/seed/components.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAUH;;;GAGG;AACH,MAAM,CAAC,MAAM,eAAe,GAAoB;IAC9C,+EAA+E;IAE/E;QACE,IAAI,EAAE,uBAAuB;QAC7B,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,4FAA4F;YAC5F,0FAA0F;KAC7F;IACD;QACE,IAAI,EAAE,wBAAwB;QAC9B,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,0FAA0F;YAC1F,wFAAwF;YACxF,yBAAyB;KAC5B;IACD;QACE,IAAI,EAAE,uBAAuB;QAC7B,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,qFAAqF;YACrF,2FAA2F;YAC3F,gDAAgD;KACnD;IACD;QACE,IAAI,EAAE,uBAAuB;QAC7B,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,2FAA2F;YAC3F,6EAA6E;KAChF;IACD;QACE,IAAI,EAAE,kBAAkB;QACxB,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,4FAA4F;YAC5F,2FAA2F;YAC3F,uFAAuF;KAC1F;IAED,+EAA+E;IAE/E;QACE,IAAI,EAAE,mBAAmB;QACzB,IAAI,EAAE,UAAU;QAChB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,iBAAiB;YACjB,2FAA2F;YAC3F,wDAAwD;YACxD,eAAe;YACf,2FAA2F;YAC3F,+DAA+D;YAC/D,4EAA4E;YAC5E,6DAA6D;YAC7D,0FAA0F;YAC1F,wFAAwF;YACxF,6FAA6F;YAC7F,6EAA6E;KAChF;IACD;QACE,IAAI,EAAE,kBAAkB;QACxB,IAAI,EAAE,UAAU;QAChB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,iBAAiB;YACjB,8FAA8F;YAC9F,wCAAwC;YACxC,eAAe;YACf,wDAAwD;YACxD,8CAA8C;YAC9C,gGAAgG;YAChG,kEAAkE;YAClE,wFAAwF;YACxF,4FAA4F;KAC/F;IAED,+EAA+E;IAE/E;QACE,IAAI,EAAE,iBAAiB;QACvB,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,kCAAkC;YAClC,uFAAuF;YACvF,0FAA0F;YAC1F,iFAAiF;YACjF,qFAAqF;YACrF,wFAAwF;YACxF,eAAe;KAClB;IACD;QACE,IAAI,EAAE,gBAAgB;QACtB,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,wFAAwF;YACxF,yFAAyF;YACzF,gFAAgF;KACnF;IACD;QACE,IAAI,EAAE,iBAAiB;QACvB,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,mFAAmF;YACnF,sFAAsF;YACtF,2FAA2F;YAC3F,gDAAgD;KACnD;IACD;QACE,IAAI,EAAE,iBAAiB;QACvB,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,sFAAsF;YACtF,wFAAwF;YACxF,wFAAwF;KAC3F;IAED,+EAA+E;IAE/E;QACE,IAAI,EAAE,eAAe;QACrB,IAAI,EAAE,SAAS;QACf,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,qBAAqB;YACrB,wFAAwF;YACxF,+EAA+E;YAC/E,2EAA2E;YAC3E,sFAAsF;YACtF,2CAA2C;YAC3C,0FAA0F;YAC1F,+DAA+D;YAC/D,sFAAsF;YACtF,mDAAmD;YACnD,uFAAuF;YACvF,mDAAmD;YACnD,0FAA0F;YAC1F,uCAAuC;KAC1C;IACD;QACE,IAAI,EAAE,kBAAkB;QACxB,IAAI,EAAE,SAAS;QACf,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,4BAA4B;YAC5B,uFAAuF;YACvF,+EAA+E;YAC/E,UAAU;YACV,uFAAuF;YACvF,gCAAgC;YAChC,uFAAuF;YACvF,wCAAwC;YACxC,iEAAiE;YACjE,yFAAyF;YACzF,qDAAqD;YACrD,uFAAuF;YACvF,qBAAqB;YACrB,uEAAuE;KAC1E;IAED,+EAA+E;IAE/E;QACE,IAAI,EAAE,aAAa;QACnB,IAAI,EAAE,SAAS;QACf,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,yBAAyB;YACzB,6FAA6F;YAC7F,eAAe;YACf,oFAAoF;YACpF,gEAAgE;YAChE,qBAAqB;YACrB,sFAAsF;YACtF,0CAA0C;YAC1C,4BAA4B;YAC5B,wFAAwF;YACxF,wFAAwF;YACxF,gBAAgB;YAChB,qFAAqF;YACrF,wFAAwF;YACxF,iBAAiB;YACjB,kBAAkB;YAClB,qFAAqF;YACrF,6BAA6B;KAChC;IAED,+EAA+E;IAE/E;QACE,IAAI,EAAE,uBAAuB;QAC7B,IAAI,EAAE,YAAY;QAClB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,0BAA0B;YAC1B,2BAA2B;YAC3B,4CAA4C;YAC5C,4EAA4E;YAC5E,6BAA6B;YAC7B,+CAA+C;YAC/C,qBAAqB;YACrB,4BAA4B;YAC5B,6BAA6B;YAC7B,yFAAyF;YACzF,qFAAqF;YACrF,4CAA4C;YAC5C,gBAAgB;YAChB,2BAA2B;YAC3B,8EAA8E;YAC9E,mBAAmB;YACnB,gCAAgC;YAChC,wFAAwF;YACxF,qFAAqF;YACrF,6BAA6B;KAChC;IAED,8EAA8E;IAE9E;QACE,IAAI,EAAE,sBAAsB;QAC5B,IAAI,EAAE,kBAAkB;QACxB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,uCAAuC;YACvC,wFAAwF;YACxF,kGAAkG;YAClG,uFAAuF;YACvF,oEAAoE;YACpE,0EAA0E;YAC1E,0CAA0C;YAC1C,sFAAsF;YACtF,oEAAoE;YACpE,iFAAiF;KACpF;IACD;QACE,IAAI,EAAE,yBAAyB;QAC/B,IAAI,EAAE,kBAAkB;QACxB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,0CAA0C;YAC1C,wFAAwF;YACxF,qFAAqF;YACrF,iEAAiE;YACjE,2FAA2F;YAC3F,+FAA+F;YAC/F,wFAAwF;YACxF,8EAA8E;YAC9E,oEAAoE;YACpE,+CAA+C;KAClD;IACD;QACE,IAAI,EAAE,0BAA0B;QAChC,IAAI,EAAE,kBAAkB;QACxB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,2CAA2C;YAC3C,qCAAqC;YACrC,mFAAmF;YACnF,oFAAoF;YACpF,4EAA4E;YAC5E,yFAAyF;YACzF,kEAAkE;YAClE,0EAA0E;KAC7E;IAED,+EAA+E;IAE/E;QACE,IAAI,EAAE,yBAAyB;QAC/B,IAAI,EAAE,YAAY;QAClB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,mBAAmB;YACnB,wFAAwF;YACxF,kFAAkF;YAClF,uBAAuB;YACvB,yDAAyD;YACzD,kEAAkE;YAClE,0EAA0E;YAC1E,+CAA+C;YAC/C,sDAAsD;YACtD,gEAAgE;YAChE,8BAA8B;YAC9B,mCAAmC;YACnC,mEAAmE;YACnE,oEAAoE;YACpE,6FAA6F;KAChG;IAED,+EAA+E;IAE/E;QACE,IAAI,EAAE,yBAAyB;QAC/B,IAAI,EAAE,aAAa;QACnB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,yCAAyC;YACzC,kBAAkB;YAClB,OAAO;YACP,uBAAuB;YACvB,wBAAwB;YACxB,mEAAmE;YACnE,wCAAwC;YACxC,OAAO;YACP,SAAS;YACT,oBAAoB;YACpB,OAAO;YACP,yBAAyB;YACzB,eAAe;YACf,0DAA0D;YAC1D,kDAAkD;YAClD,0DAA0D;YAC1D,YAAY;YACZ,4BAA4B;YAC5B,kDAAkD;YAClD,KAAK;KACR;IACD;QACE,IAAI,EAAE,gCAAgC;QACtC,IAAI,EAAE,aAAa;QACnB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,iDAAiD;YACjD,OAAO;YACP,0DAA0D;YAC1D,uBAAuB;YACvB,wDAAwD;YACxD,iEAAiE;YACjE,wEAAwE;YACxE,4EAA4E;YAC5E,sBAAsB;YACtB,iGAAiG;YACjG,SAAS;YACT,sFAAsF;YACtF,wFAAwF;KAC3F;IAED,8EAA8E;IAE9E;QACE,IAAI,EAAE,0BAA0B;QAChC,IAAI,EAAE,UAAU;QAChB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,sDAAsD;YACtD,yBAAyB;YACzB,yEAAyE;YACzE,yCAAyC;YACzC,iFAAiF;YACjF,0EAA0E;YAC1E,0BAA0B;YAC1B,2BAA2B;YAC3B,2BAA2B;YAC3B,oDAAoD;YACpD,sDAAsD;YACtD,wBAAwB;YACxB,gFAAgF;YAChF,4DAA4D;KAC/D;IAED,+EAA+E;IAE/E;QACE,IAAI,EAAE,yBAAyB;QAC/B,IAAI,EAAE,aAAa;QACnB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,+FAA+F;YAC/F,kBAAkB;YAClB,0DAA0D;YAC1D,wCAAwC;YACxC,8CAA8C;YAC9C,gFAAgF;YAChF,6DAA6D;YAC7D,oDAAoD;YACpD,2BAA2B;YAC3B,yFAAyF;YACzF,uCAAuC;YACvC,wBAAwB;YACxB,qFAAqF;KACxF;IACD;QACE,IAAI,EAAE,mBAAmB;QACzB,IAAI,EAAE,UAAU;QAChB,OAAO,EAAE,CAAC;QACV,QAAQ,EAAE,IAAI;QACd,OAAO,EACL,2BAA2B;YAC3B,iGAAiG;YACjG,qDAAqD;YACrD,4EAA4E;YAC5E,yDAAyD;YACzD,yEAAyE;KAC5E;CACF,CAAC"}

package/src/seed/index.d.ts

@@ -0,0 +1,30 @@
+/**
+ * seed/index.ts
+ *
+ * Idempotent seed function for @adhd/agent-registry.
+ *
+ * Inserts all system prompt types and shared components into the given DB.
+ * Running `seed(db)` twice is a NO-OP using two different idempotency strategies:
+ *   - Prompt types: ComponentStore.upsertType uses ON CONFLICT DO NOTHING.
+ *   - Components: read-before-write (SELECT to check existence, conditional INSERT),
+ *     not INSERT OR IGNORE — so row counts and versions are never bumped on re-seed.
+ *
+ * [inv:version-retained] — seed never calls ComponentStore.version(); it only
+ * inserts rows that are absent (read-before-write for components, ON CONFLICT DO
+ * NOTHING for prompt types).
+ *
+ * Usage:
+ *   import { seed } from '@adhd/agent-registry/seed';
+ *   seed(db);  // safe to call on every startup
+ */
+import type { BetterSQLite3Database } from 'drizzle-orm/better-sqlite3';
+/**
+ * Seed all system prompt types and shared components.
+ *
+ * Idempotent: safe to call on every application startup.
+ * Second and subsequent calls are no-ops for all rows already present.
+ *
+ * @param db - An open Drizzle BetterSQLite3 database with migrations already run.
+ */
+export declare function seed(db: BetterSQLite3Database<any>): void;
+//# sourceMappingURL=index.d.ts.map

package/src/seed/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../../../packages/agent/agent-store-prompts/src/seed/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAGH,OAAO,KAAK,EAAE,qBAAqB,EAAE,MAAM,4BAA4B,CAAC;AAQxE;;;;;;;GAOG;AAEH,wBAAgB,IAAI,CAAC,EAAE,EAAE,qBAAqB,CAAC,GAAG,CAAC,GAAG,IAAI,CAwEzD"}

package/src/seed/index.js

@@ -0,0 +1,95 @@
+/**
+ * seed/index.ts
+ *
+ * Idempotent seed function for @adhd/agent-registry.
+ *
+ * Inserts all system prompt types and shared components into the given DB.
+ * Running `seed(db)` twice is a NO-OP using two different idempotency strategies:
+ *   - Prompt types: ComponentStore.upsertType uses ON CONFLICT DO NOTHING.
+ *   - Components: read-before-write (SELECT to check existence, conditional INSERT),
+ *     not INSERT OR IGNORE — so row counts and versions are never bumped on re-seed.
+ *
+ * [inv:version-retained] — seed never calls ComponentStore.version(); it only
+ * inserts rows that are absent (read-before-write for components, ON CONFLICT DO
+ * NOTHING for prompt types).
+ *
+ * Usage:
+ *   import { seed } from '@adhd/agent-registry/seed';
+ *   seed(db);  // safe to call on every startup
+ */
+import { and, eq } from 'drizzle-orm';
+import { ComponentStore } from '../store/component-store.js';
+import { PROMPT_TYPES } from './prompt-types.js';
+import { SEED_COMPONENTS } from './components.js';
+import { componentsTable, componentVersionsTable } from '../db/schema.js';
+/**
+ * Seed all system prompt types and shared components.
+ *
+ * Idempotent: safe to call on every application startup.
+ * Second and subsequent calls are no-ops for all rows already present.
+ *
+ * @param db - An open Drizzle BetterSQLite3 database with migrations already run.
+ */
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+export function seed(db) {
+    const store = new ComponentStore(db);
+    // ── 1. Prompt types ────────────────────────────────────────────────────────
+    //
+    // ComponentStore.upsertType uses ON CONFLICT DO NOTHING on the TEXT PRIMARY KEY
+    // of registry_prompt_types.slug, so this is genuinely idempotent.
+    for (const type of PROMPT_TYPES) {
+        store.upsertType(type);
+    }
+    // ── 2. Prompt components (head/version split, Decision 5) ───────────────────
+    //
+    // Each seed entry materializes as a registry_components head identity row plus a
+    // registry_component_versions history row at the canonical seed version.
+    //
+    // Idempotency without overwriting (read-before-write per table):
+    //   - head row: insert only if the slug has no registry_components row yet.
+    //   - version row: insert only if that exact (slug, version) is absent. The real
+    //     UNIQUE(slug, version) index would otherwise reject a duplicate, but we skip
+    //     proactively so re-seed is a clean no-op (never bumps version).
+    //
+    // [inv:version-retained] — seed never calls ComponentStore.version(); it only
+    // inserts rows that are absent.
+    const now = new Date().toISOString();
+    for (const component of SEED_COMPONENTS) {
+        // Head identity row — insert once per slug.
+        const head = db
+            .select({ slug: componentsTable.slug })
+            .from(componentsTable)
+            .where(eq(componentsTable.slug, component.slug))
+            .get();
+        if (!head) {
+            db.insert(componentsTable)
+                .values({
+                slug: component.slug,
+                type: component.type,
+                isShared: component.isShared,
+                createdAt: now,
+            })
+                .run();
+        }
+        // Version history row — insert only if this exact (slug, version) is absent.
+        const existingVersion = db
+            .select({ versionId: componentVersionsTable.versionId })
+            .from(componentVersionsTable)
+            .where(and(eq(componentVersionsTable.slug, component.slug), eq(componentVersionsTable.version, component.version)))
+            .get();
+        if (existingVersion) {
+            // Row already seeded — skip; never overwrite content on re-seed.
+            continue;
+        }
+        db.insert(componentVersionsTable)
+            .values({
+            slug: component.slug,
+            version: component.version,
+            content: component.content,
+            createdAt: now,
+            updatedAt: now,
+        })
+            .run();
+    }
+}
+//# sourceMappingURL=index.js.map

package/src/seed/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../../../../packages/agent/agent-store-prompts/src/seed/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAEH,OAAO,EAAE,GAAG,EAAE,EAAE,EAAE,MAAM,aAAa,CAAC;AAGtC,OAAO,EAAE,cAAc,EAAE,MAAM,6BAA6B,CAAC;AAC7D,OAAO,EAAE,YAAY,EAAE,MAAM,mBAAmB,CAAC;AACjD,OAAO,EAAE,eAAe,EAAE,MAAM,iBAAiB,CAAC;AAElD,OAAO,EAAE,eAAe,EAAE,sBAAsB,EAAE,MAAM,iBAAiB,CAAC;AAE1E;;;;;;;GAOG;AACH,8DAA8D;AAC9D,MAAM,UAAU,IAAI,CAAC,EAA8B;IACjD,MAAM,KAAK,GAAG,IAAI,cAAc,CAAC,EAAE,CAAC,CAAC;IAErC,8EAA8E;IAC9E,EAAE;IACF,gFAAgF;IAChF,kEAAkE;IAClE,KAAK,MAAM,IAAI,IAAI,YAAY,EAAE,CAAC;QAChC,KAAK,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC;IACzB,CAAC;IAED,+EAA+E;IAC/E,EAAE;IACF,iFAAiF;IACjF,yEAAyE;IACzE,EAAE;IACF,iEAAiE;IACjE,4EAA4E;IAC5E,iFAAiF;IACjF,kFAAkF;IAClF,qEAAqE;IACrE,EAAE;IACF,8EAA8E;IAC9E,gCAAgC;IAChC,MAAM,GAAG,GAAG,IAAI,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;IAErC,KAAK,MAAM,SAAS,IAAI,eAAe,EAAE,CAAC;QACxC,4CAA4C;QAC5C,MAAM,IAAI,GAAG,EAAE;aACZ,MAAM,CAAC,EAAE,IAAI,EAAE,eAAe,CAAC,IAAI,EAAE,CAAC;aACtC,IAAI,CAAC,eAAe,CAAC;aACrB,KAAK,CAAC,EAAE,CAAC,eAAe,CAAC,IAAI,EAAE,SAAS,CAAC,IAAI,CAAC,CAAC;aAC/C,GAAG,EAAE,CAAC;QAET,IAAI,CAAC,IAAI,EAAE,CAAC;YACV,EAAE,CAAC,MAAM,CAAC,eAAe,CAAC;iBACvB,MAAM,CAAC;gBACN,IAAI,EAAE,SAAS,CAAC,IAAI;gBACpB,IAAI,EAAE,SAAS,CAAC,IAAI;gBACpB,QAAQ,EAAE,SAAS,CAAC,QAAQ;gBAC5B,SAAS,EAAE,GAAG;aACf,CAAC;iBACD,GAAG,EAAE,CAAC;QACX,CAAC;QAED,6EAA6E;QAC7E,MAAM,eAAe,GAAG,EAAE;aACvB,MAAM,CAAC,EAAE,SAAS,EAAE,sBAAsB,CAAC,SAAS,EAAE,CAAC;aACvD,IAAI,CAAC,sBAAsB,CAAC;aAC5B,KAAK,CACJ,GAAG,CACD,EAAE,CAAC,sBAAsB,CAAC,IAAI,EAAE,SAAS,CAAC,IAAI,CAAC,EAC/C,EAAE,CAAC,sBAAsB,CAAC,OAAO,EAAE,SAAS,CAAC,OAAO,CAAC,CACtD,CACF;aACA,GAAG,EAAE,CAAC;QAET,IAAI,eAAe,EAAE,CAAC;YACpB,iEAAiE;YACjE,SAAS;QACX,CAAC;QAED,EAAE,CAAC,MAAM,CAAC,sBAAsB,CAAC;aAC9B,MAAM,CAAC;YACN,IAAI,EAAE,SAAS,CAAC,IAAI;YACpB,OAAO,EAAE,SAAS,CAAC,OAAO;YAC1B,OAAO,EAAE,SAAS,CAAC,OAAO;YAC1B,SAAS,EAAE,GAAG;YACd,SAAS,EAAE,GAAG;SACf,CAAC;aACD,GAAG,EAAE,CAAC;IACX,CAAC;AACH,CAAC"}

package/src/seed/prompt-types.d.ts

@@ -0,0 +1,17 @@
+/**
+ * seed/prompt-types.ts
+ *
+ * Every system prompt type that ships with @adhd/agent-registry.
+ * These map 1:1 to rows in registry_prompt_types (is_system = true).
+ *
+ * Source: docs/plan/agent-registry/SEED_DATA.md §1
+ * [inv:lookup-not-enum] — types are rows, not SQL enums. Adding a new type
+ * is a seed operation, not a migration.
+ */
+import type { PromptType } from '../store/component-store.js';
+/**
+ * All system prompt types for the registry.
+ * Array ordering is cosmetic — types are keyed by slug in the DB.
+ */
+export declare const PROMPT_TYPES: PromptType[];
+//# sourceMappingURL=prompt-types.d.ts.map

package/src/seed/prompt-types.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"prompt-types.d.ts","sourceRoot":"","sources":["../../../../../../packages/agent/agent-store-prompts/src/seed/prompt-types.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AAEH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE9D;;;GAGG;AACH,eAAO,MAAM,YAAY,EAAE,UAAU,EAgGpC,CAAC"}