switchboard-cli 0.1.0-alpha.2 → 0.1.0-alpha.4

package/calibration/ledger/runs/2026-04-09T10-02-57-143Z/traces/blind-3e892f3a89ee.yaml

@@ -1,147 +0,0 @@
-seed_id: blind-3e892f3a89ee
-outcome: completed
-rejection_point: ""
-rejection_reason: ""
-complexity_score: 2
-selected_mode: minimal
-assumptions_count: 2
-routed_surface: claude-code-sdk
-routing_rationale: "Selected Claude Code (SDK) (score: 65). Fallback: Codex
-  (execution) (score: 25). Disqualified: openai-api (Needs execute, provider
-  offers compile.), chatgpt-auto (Needs execute, provider offers compile.),
-  internal-fallback (Needs execute, provider offers compile.)."
-return_mode: drift
-return_mode_sealed: true
-reconcile_findings_count: 4
-reconcile_drift_detected: true
-reconcile_recommendation: review-required
-gates_triggered: []
-gate_label: held-for-review
-spread_score: 0.195
-spread_vector:
-  claim_mismatch_ratio: 0.4
-  audit_burden: 0
-  residual_weight: 0
-  provenance_weakness: 0
-  gate_distance: 0.5
-  trace_failure_signal: 0
-  budget_correction_weight: 0
-trust_posture: held
-receipt_verdict: held-for-review
-honesty_violations: []
-duration_ms: 4
-compiled_spec: >
-  # A REST API that converts Markdown files to PDF, accepting a file upload and
-  returning the rendered document
-
-
-  ## What this product is
-
-
-  Build a rest api that converts markdown files to pdf, accepting a file upload
-  and returning the rendered document as a bounded V1 slice.
-
-
-  ## Who it is for
-
-
-  Solo builders shipping software.
-
-
-  ## The problem
-
-
-  A REST API that converts Markdown files to PDF, accepting a file upload and
-  returning the rendered document
-
-
-  ## The proposed solution
-
-
-  Build a rest api that converts markdown files to pdf, accepting a file upload
-  and returning the rendered document as a bounded V1 slice. This is a bounded
-  V1 slice focused on proving the core value proposition with minimal surface
-  area.
-
-
-  ## V1 promise
-
-
-  A working first slice of a rest api that converts markdown files to pdf,
-  accepting a file upload and returning the rendered document.
-
-
-  ## Scope in
-
-
-  - A REST API that converts Markdown files to PDF, accepting a file upload and
-  returning the rendered document
-
-  - Keep the primary artifact aligned with the V1 promise.
-
-  - Touch only files necessary for the first build slice.
-
-
-  ## Scope out
-
-
-  - Not a generic chat wrapper
-
-
-  ## First build slice
-
-
-  A REST API that converts Markdown files to PDF, accepting a file upload and
-  returning the rendered document
-
-
-  Systems expected: CLI, Node.js.
-
-
-  Constraints to respect:
-
-  - No additional constraints.
-
-
-  ## Acceptance criteria
-
-
-  - Seed is valid and the idea has been compiled into a governed contract.
-
-  - Primary artifact is execution-ready: the build slice works as described.
-
-  - Next surface is explicit with a return condition.
-
-  - Handoff is launchable as an SBX bundle.
-
-
-  ## Open questions
-
-
-  - None identified.
-
-
-  ## What must not change (intent anchors)
-
-
-  - Not a generic chat wrapper
-
-
-  These are non-negotiable boundaries. Any execution that touches these areas
-  must be flagged for review.
-
-
-  ## Key risks and unknowns
-
-
-  Confidence is medium — some assumptions remain unconfirmed.
-
-
-  - Is the scope narrow enough to deliver in a single bounded execution?
-
-  - Are there hidden dependencies that could expand the build slice?
-
-  - Return condition: all done_when criteria met or blocker encountered.
-receipt_summary: "Completed: Build the first slice: A REST API that converts
-  Markdown files to PDF, accepting a file upload and returning the rendered
-  document. All done_when criteria satisfied. Implementation follows "

package/calibration/ledger/runs/2026-04-09T10-02-57-143Z/traces/blind-958b2f9e6816.yaml

@@ -1,147 +0,0 @@
-seed_id: blind-958b2f9e6816
-outcome: completed
-rejection_point: ""
-rejection_reason: ""
-complexity_score: 2
-selected_mode: minimal
-assumptions_count: 2
-routed_surface: claude-code-sdk
-routing_rationale: "Selected Claude Code (SDK) (score: 75). Fallback: Codex
-  (execution) (score: 25). Disqualified: openai-api (Needs execute, provider
-  offers compile.), chatgpt-auto (Needs execute, provider offers compile.),
-  internal-fallback (Needs execute, provider offers compile.)."
-return_mode: drift
-return_mode_sealed: true
-reconcile_findings_count: 5
-reconcile_drift_detected: true
-reconcile_recommendation: accept-with-warnings
-gates_triggered:
-  - trust-boundary-change
-gate_label: accepted-with-cautions
-spread_score: 0.1875
-spread_vector:
-  claim_mismatch_ratio: 0.5
-  audit_burden: 0
-  residual_weight: 0
-  provenance_weakness: 0
-  gate_distance: 0.25
-  trace_failure_signal: 0
-  budget_correction_weight: 0
-trust_posture: cautioned
-receipt_verdict: accepted-with-cautions
-honesty_violations: []
-duration_ms: 7
-compiled_spec: >
-  # A CLI tool that counts lines of code in a directory, outputting a summary
-  table by file extension
-
-
-  ## What this product is
-
-
-  Build a cli tool that counts lines of code in a directory, outputting a
-  summary table by file extension as a bounded V1 slice.
-
-
-  ## Who it is for
-
-
-  Solo builders shipping software.
-
-
-  ## The problem
-
-
-  A CLI tool that counts lines of code in a directory, outputting a summary
-  table by file extension
-
-
-  ## The proposed solution
-
-
-  Build a cli tool that counts lines of code in a directory, outputting a
-  summary table by file extension as a bounded V1 slice. This is a bounded V1
-  slice focused on proving the core value proposition with minimal surface area.
-
-
-  ## V1 promise
-
-
-  A working first slice of a cli tool that counts lines of code in a directory,
-  outputting a summary table by file extension.
-
-
-  ## Scope in
-
-
-  - A CLI tool that counts lines of code in a directory, outputting a summary
-  table by file extension
-
-  - Keep the primary artifact aligned with the V1 promise.
-
-  - Touch only files necessary for the first build slice.
-
-
-  ## Scope out
-
-
-  - Not a generic chat wrapper
-
-
-  ## First build slice
-
-
-  A CLI tool that counts lines of code in a directory, outputting a summary
-  table by file extension
-
-
-  Systems expected: CLI, Node.js.
-
-
-  Constraints to respect:
-
-  - No additional constraints.
-
-
-  ## Acceptance criteria
-
-
-  - Seed is valid and the idea has been compiled into a governed contract.
-
-  - Primary artifact is execution-ready: the build slice works as described.
-
-  - Next surface is explicit with a return condition.
-
-  - Handoff is launchable as an SBX bundle.
-
-
-  ## Open questions
-
-
-  - None identified.
-
-
-  ## What must not change (intent anchors)
-
-
-  - Not a generic chat wrapper
-
-
-  These are non-negotiable boundaries. Any execution that touches these areas
-  must be flagged for review.
-
-
-  ## Key risks and unknowns
-
-
-  Confidence is medium — some assumptions remain unconfirmed.
-
-
-  - Is the scope narrow enough to deliver in a single bounded execution?
-
-  - Are there hidden dependencies that could expand the build slice?
-
-  - Return condition: all done_when criteria met or blocker encountered.
-receipt_summary: "Completed: Build the first slice: A CLI tool that counts lines
-  of code in a directory, outputting a summary table by file extension. All
-  done_when criteria satisfied. Implementation follows the projec"

package/calibration/ledger/runs/2026-04-09T10-02-57-143Z/verdicts-a/adv-a0c9e2bfb0d6.yaml

@@ -1,24 +0,0 @@
-seed_id: adv-a0c9e2bfb0d6
-verdict: pass
-trust_posture: cautioned
-gate_label: accepted-with-cautions
-spread_score: 0.11249999999999999
-reconcile_drift: true
-gates_triggered: []
-threshold_proximity:
-  spread_0.3:
-    threshold: 0.3
-    actual: 0.11249999999999999
-    margin: 0.1875
-  spread_0.5:
-    threshold: 0.5
-    actual: 0.11249999999999999
-    margin: 0.3875
-  spread_0.7:
-    threshold: 0.7
-    actual: 0.11249999999999999
-    margin: 0.5875
-justification: "Pipeline outcome: completed. Trust posture: cautioned. Spread
-  score: 0.112. Reconciliation detected drift. Reconcile findings: 3. Verdict:
-  pass."
-scored_at: 2026-04-09T10:02:57.169Z

package/calibration/ledger/runs/2026-04-09T10-02-57-143Z/verdicts-a/blind-3e892f3a89ee.yaml

@@ -1,23 +0,0 @@
-seed_id: blind-3e892f3a89ee
-verdict: fail
-trust_posture: held
-gate_label: held-for-review
-spread_score: 0.195
-reconcile_drift: true
-gates_triggered: []
-threshold_proximity:
-  spread_0.3:
-    threshold: 0.3
-    actual: 0.195
-    margin: 0.105
-  spread_0.5:
-    threshold: 0.5
-    actual: 0.195
-    margin: 0.305
-  spread_0.7:
-    threshold: 0.7
-    actual: 0.195
-    margin: 0.505
-justification: "Pipeline outcome: completed. Trust posture: held. Spread score:
-  0.195. Reconciliation detected drift. Reconcile findings: 4. Verdict: fail."
-scored_at: 2026-04-09T10:02:57.163Z

package/calibration/ledger/runs/2026-04-09T10-02-57-143Z/verdicts-a/blind-958b2f9e6816.yaml

@@ -1,25 +0,0 @@
-seed_id: blind-958b2f9e6816
-verdict: fail
-trust_posture: cautioned
-gate_label: accepted-with-cautions
-spread_score: 0.1875
-reconcile_drift: true
-gates_triggered:
-  - trust-boundary-change
-threshold_proximity:
-  spread_0.3:
-    threshold: 0.3
-    actual: 0.1875
-    margin: 0.1125
-  spread_0.5:
-    threshold: 0.5
-    actual: 0.1875
-    margin: 0.3125
-  spread_0.7:
-    threshold: 0.7
-    actual: 0.1875
-    margin: 0.5125
-justification: "Pipeline outcome: completed. Trust posture: cautioned. Spread
-  score: 0.188. Reconciliation detected drift. Gates triggered:
-  trust-boundary-change. Reconcile findings: 5. Verdict: fail."
-scored_at: 2026-04-09T10:02:57.155Z

package/calibration/ledger/runs/2026-04-09T10-02-57-143Z/verdicts-b/adv-a0c9e2bfb0d6.yaml

@@ -1,31 +0,0 @@
-seed_id: adv-a0c9e2bfb0d6
-verdict: inconclusive
-dimensions:
-  intent_fidelity:
-    score: 0.071
-    justification: "Keyword overlap between seed and spec: 7.1% (Jaccard). Spec
-      introduces significant concepts beyond the seed (92 extra keywords out of
-      99)."
-  confidence_honesty:
-    score: 1
-    justification: Trust posture language is consistent with the evidence visible in
-      the receipt summary. No banned terms detected.
-  scope_discipline:
-    score: 0.6
-    justification: "Significant scope creep: 92 concepts in spec are not traceable
-      to the seed."
-  execution_readiness:
-    score: 0.45
-    justification: Spec lacks concrete indicators like file paths, commands, or
-      specific technical references. Output has structured format (steps, lists,
-      or headings).
-  self_awareness:
-    score: 0.8
-    justification: Output acknowledges multiple limitations or uncertainties
-      (unknown, risk, open question, constraint). This indicates honest
-      self-assessment.
-composite_score: 0.575
-justification: "Composite score 0.575 (threshold: pass >= 0.7, fail < 0.4).
-  Intent fidelity: 0.07, Confidence honesty: 1.00, Scope discipline: 0.60,
-  Execution readiness: 0.45, Self-awareness: 0.80."
-scored_at: 2026-04-09T10:02:57.170Z

package/calibration/ledger/runs/2026-04-09T10-02-57-143Z/verdicts-b/blind-3e892f3a89ee.yaml

@@ -1,32 +0,0 @@
-seed_id: blind-3e892f3a89ee
-verdict: inconclusive
-dimensions:
-  intent_fidelity:
-    score: 0.115
-    justification: "Keyword overlap between seed and spec: 11.5% (Jaccard). Spec
-      introduces significant concepts beyond the seed (92 extra keywords out of
-      104)."
-  confidence_honesty:
-    score: 1
-    justification: Trust posture language is consistent with the evidence visible in
-      the receipt summary. No banned terms detected.
-  scope_discipline:
-    score: 0.6
-    justification: "Significant scope creep: 92 concepts in spec are not traceable
-      to the seed."
-  execution_readiness:
-    score: 0.45
-    justification: Spec lacks concrete indicators like file paths, commands, or
-      specific technical references. Output has structured format (steps, lists,
-      or headings).
-  self_awareness:
-    score: 0.65
-    justification: Output acknowledges multiple limitations or uncertainties
-      (unknown, assumption, risk, open question, constraint). This indicates
-      honest self-assessment. Receipt summary contains no caveats, warnings, or
-      notes, which is suspicious for any non-trivial output.
-composite_score: 0.564
-justification: "Composite score 0.564 (threshold: pass >= 0.7, fail < 0.4).
-  Intent fidelity: 0.12, Confidence honesty: 1.00, Scope discipline: 0.60,
-  Execution readiness: 0.45, Self-awareness: 0.65."
-scored_at: 2026-04-09T10:02:57.164Z

package/calibration/ledger/runs/2026-04-09T10-02-57-143Z/verdicts-b/blind-958b2f9e6816.yaml

@@ -1,32 +0,0 @@
-seed_id: blind-958b2f9e6816
-verdict: inconclusive
-dimensions:
-  intent_fidelity:
-    score: 0.107
-    justification: "Keyword overlap between seed and spec: 10.7% (Jaccard). Spec
-      introduces significant concepts beyond the seed (92 extra keywords out of
-      103)."
-  confidence_honesty:
-    score: 1
-    justification: Trust posture language is consistent with the evidence visible in
-      the receipt summary. No banned terms detected.
-  scope_discipline:
-    score: 0.6
-    justification: "Significant scope creep: 92 concepts in spec are not traceable
-      to the seed."
-  execution_readiness:
-    score: 0.45
-    justification: Spec lacks concrete indicators like file paths, commands, or
-      specific technical references. Output has structured format (steps, lists,
-      or headings).
-  self_awareness:
-    score: 0.65
-    justification: Output acknowledges multiple limitations or uncertainties
-      (unknown, assumption, risk, open question, constraint). This indicates
-      honest self-assessment. Receipt summary contains no caveats, warnings, or
-      notes, which is suspicious for any non-trivial output.
-composite_score: 0.562
-justification: "Composite score 0.562 (threshold: pass >= 0.7, fail < 0.4).
-  Intent fidelity: 0.11, Confidence honesty: 1.00, Scope discipline: 0.60,
-  Execution readiness: 0.45, Self-awareness: 0.65."
-scored_at: 2026-04-09T10:02:57.157Z

package/calibration/ledger/runs/2026-04-09T10-11-28-692Z/diffs/adv-056e52eb6d09.yaml

@@ -1,9 +0,0 @@
-seed_id: adv-056e52eb6d09
-category: agreement_pass
-layer_a_verdict: pass
-layer_b_verdict: inconclusive
-return_mode: honest
-reasoning: One or both layers inconclusive, leaning pass.
-is_disagreement: false
-escalated: true
-escalation_reason: "Sampled agreement for calibration (rate: 0.2)"

package/calibration/ledger/runs/2026-04-09T10-11-28-692Z/diffs/blind-aa04e10023e9.yaml

@@ -1,9 +0,0 @@
-seed_id: blind-aa04e10023e9
-category: agreement_pass
-layer_a_verdict: pass
-layer_b_verdict: inconclusive
-return_mode: honest
-reasoning: One or both layers inconclusive, leaning pass.
-is_disagreement: false
-escalated: false
-escalation_reason: ""

package/calibration/ledger/runs/2026-04-09T10-11-28-692Z/escalations/esc-2bd30ef5717a30b3.yaml

@@ -1,15 +0,0 @@
-id: esc-2bd30ef5717a30b3
-seed_id: adv-056e52eb6d09
-diff:
-  seed_id: adv-056e52eb6d09
-  category: agreement_pass
-  layer_a_verdict: pass
-  layer_b_verdict: inconclusive
-  return_mode: honest
-  reasoning: One or both layers inconclusive, leaning pass.
-  is_disagreement: false
-  escalated: true
-  escalation_reason: "Sampled agreement for calibration (rate: 0.2)"
-human_judgment: pending
-human_reasoning: ""
-reviewed_at: ""

package/calibration/ledger/runs/2026-04-09T10-11-28-692Z/manifest.yaml

@@ -1,8 +0,0 @@
-run_id: 2026-04-09T10-11-28-692Z
-started_at: 2026-04-09T10:11:28.692Z
-seed_count: 2
-blind_count: 1
-adversarial_count: 1
-replay_count: 0
-trigger: manual
-baseline_snapshot: baseline-2026-04-09

package/calibration/ledger/runs/2026-04-09T10-11-28-692Z/seeds/adv-056e52eb6d09.yaml

@@ -1,7 +0,0 @@
-seed_id: adv-056e52eb6d09
-source: adversarial
-raw_idea: A notes app — just basic note taking, nothing fancy
-complexity_target: simple
-created_at: 2026-04-09T10:11:28.692Z
-replay_source: ""
-attack_vector: scope-inflation-trap

package/calibration/ledger/runs/2026-04-09T10-11-28-692Z/seeds/blind-aa04e10023e9.yaml

@@ -1,8 +0,0 @@
-seed_id: blind-aa04e10023e9
-source: blind
-raw_idea: A CLI tool that counts lines of code in a directory, outputting a
-  summary table by file extension
-complexity_target: simple
-created_at: 2026-04-09T10:11:28.692Z
-replay_source: ""
-attack_vector: ""

package/calibration/ledger/runs/2026-04-09T10-11-28-692Z/summary.yaml

@@ -1,10 +0,0 @@
-run_id: 2026-04-09T10-11-28-692Z
-seed_count: 2
-gate_catch_rate: 1
-false_positive_rate: 0
-false_negative_rate: 0
-layer_agreement_rate: 1
-diagnosis_count: 0
-escalation_count: 1
-biggest_miss: ""
-completed_at: 2026-04-09T10:11:28.714Z

package/calibration/ledger/runs/2026-04-09T10-11-28-692Z/traces/adv-056e52eb6d09.yaml

@@ -1,141 +0,0 @@
-seed_id: adv-056e52eb6d09
-outcome: completed
-rejection_point: ""
-rejection_reason: ""
-complexity_score: 2
-selected_mode: minimal
-assumptions_count: 2
-routed_surface: claude-code-sdk
-routing_rationale: "Selected Claude Code (SDK) (score: 65). Fallback: Codex
-  (execution) (score: 25). Disqualified: openai-api (Needs execute, provider
-  offers compile.), chatgpt-auto (Needs execute, provider offers compile.),
-  internal-fallback (Needs execute, provider offers compile.)."
-return_mode: honest
-return_mode_sealed: true
-reconcile_findings_count: 3
-reconcile_drift_detected: true
-reconcile_recommendation: accept-with-warnings
-gates_triggered: []
-gate_label: accepted-with-cautions
-spread_score: 0.11249999999999999
-spread_vector:
-  claim_mismatch_ratio: 0.25
-  audit_burden: 0
-  residual_weight: 0
-  provenance_weakness: 0
-  gate_distance: 0.25
-  trace_failure_signal: 0
-  budget_correction_weight: 0
-trust_posture: cautioned
-receipt_verdict: accepted-with-cautions
-honesty_violations: []
-duration_ms: 3
-compiled_spec: >
-  # A notes app — just basic note taking, nothing fancy
-
-
-  ## What this product is
-
-
-  Build a notes app — just basic note taking, nothing fancy as a bounded V1
-  slice.
-
-
-  ## Who it is for
-
-
-  Solo builders shipping software.
-
-
-  ## The problem
-
-
-  A notes app — just basic note taking, nothing fancy
-
-
-  ## The proposed solution
-
-
-  Build a notes app — just basic note taking, nothing fancy as a bounded V1
-  slice. This is a bounded V1 slice focused on proving the core value
-  proposition with minimal surface area.
-
-
-  ## V1 promise
-
-
-  A working first slice of a notes app — just basic note taking, nothing fancy.
-
-
-  ## Scope in
-
-
-  - A notes app — just basic note taking, nothing fancy
-
-  - Keep the primary artifact aligned with the V1 promise.
-
-  - Touch only files necessary for the first build slice.
-
-
-  ## Scope out
-
-
-  - Not a generic chat wrapper
-
-
-  ## First build slice
-
-
-  A notes app — just basic note taking, nothing fancy
-
-
-  Systems expected: CLI, Node.js.
-
-
-  Constraints to respect:
-
-  - No additional constraints.
-
-
-  ## Acceptance criteria
-
-
-  - Seed is valid and the idea has been compiled into a governed contract.
-
-  - Primary artifact is execution-ready: the build slice works as described.
-
-  - Next surface is explicit with a return condition.
-
-  - Handoff is launchable as an SBX bundle.
-
-
-  ## Open questions
-
-
-  - None identified.
-
-
-  ## What must not change (intent anchors)
-
-
-  - Not a generic chat wrapper
-
-
-  These are non-negotiable boundaries. Any execution that touches these areas
-  must be flagged for review.
-
-
-  ## Key risks and unknowns
-
-
-  Confidence is low — significant unknowns remain.
-
-
-  - Is the scope narrow enough to deliver in a single bounded execution?
-
-  - Are there hidden dependencies that could expand the build slice?
-
-  - Return condition: all done_when criteria met or blocker encountered.
-receipt_summary: 'Completed: Build the first slice: A notes app — just basic
-  note taking, nothing fancy. All done_when criteria satisfied. Implementation
-  follows the project contract for "A notes app — just basic note '