Tasks: retire the execution log

The last spec artifact (requirements → design → testing plan → tasks). A DAG of implementation tasks derived from the approved design and testing plan.

Task list

• [x] 1. Bundle the evidence templates that replace the log's gated sections

  • skills/the-loop/templates/evidence/: self-review.md, critic-review.md, security-review.md, design-critic-review.md, verification.md, final-validation.md, documentation.md, pull-requests.md
  • Each carries exactly the section(s) its gate demands and nothing else — no phase table, no progress entries, no front-matter phase mirror
  • Depends on: none
  • Requirements: R2.1, R1.2
  • Test: T3 — test_p3_every_gated_name_has_a_template_that_can_satisfy_it (red→green)

• [x] 2. Re-target the five shipped graphs

  • pdlc-work-item-loop, pdlc-pr-loop, pdlc-contribution-loop: every validates: execution-log.md becomes the mapped evidence/*.md — the verb is unchanged, only the target moves (design D1)
  • Remove log-entry from all 47 entry chains across all five graphs
  • Depends on: 1
  • Requirements: R1.4, R2.1, R2.3
  • Test: T1, T3, T4 (red→green)

• [x] 3. Delete the log-entry hook

  • cli/the_loop/graph/hooks/sideeffects.py; update the docstring in runtime.py that lists the best-effort hooks
  • Depends on: 2
  • Requirements: R1.4
  • Test: T1 (red→green)

• [x] 4. Move repos: to tasks.md

  • hooks/loops.py:declared_repos reads tasks.md's front matter; the docstring states the absence rule unchanged
  • Depends on: none
  • Requirements: R3.1, R3.2, R3.3
  • Test: T2, T7 (red→green)

• [x] 5. Re-seed the dead-session fallback

  • runtime.py:resolve_session seeds requirements.md, design.md, tasks.md
  • Depends on: none
  • Requirements: R4.2
  • Test: T10 (red→green)

• [x] 6. Retire the artifact from the manifest and the template tree

  • .the-loop/manifest.yaml: drop the execution-log role, add the eight evidence entries; skills/the-loop/templates/execution-log.md deleted
  • Widen _SPEC_FILE in test_graph_parity.py so a pathPattern with one directory segment (evidence/x.md) is matched rather than silently excluded
  • Depends on: 1, 2
  • Requirements: R1.1, R1.2, R2.5
  • Test: T3 (red→green)

• [x] 7. Update the tests that assert the old shape

  • test_graph_hooks, test_graph_loops, test_graph_skips, test_graph_cleanup, test_graph_review_chain_integration, test_graph_verification_integration, test_graph_multirepo_integration, test_graphlink_integration, test_bus_integration, test_writing_parity
  • New: the skip-isolation case (T5), the legacy-log-on-disk case (T8)
  • Depends on: 2, 3, 4, 5, 6
  • Requirements: R2.3, R5.3
  • Test: T4, T5, T6, T7, T8 (red→green)

• [x] 8. Update the e2e scenario fixtures

  • cli/tests/test_pdlc_e2e/: runner.py, the three scenarios' scenario.yaml and artifacts/ — execution-log.md fixtures become the evidence/*.md the loop now gates
  • Depends on: 2, 6
  • Requirements: A7
  • Test: T9 (red→green)

• [x] 9. Rewrite the operating model

  • skills/the-loop/SKILL.md and reference/{workflow,context,reviewing,security, token-economy,tooling,instructions,automation,design-artifacts}.md
  • reference/context.md carries the new reset protocol (D4): checkmarks + commit, no prose checkpoint
  • Depends on: 2
  • Requirements: R1.5, R4.1, R4.3
  • Test: T18, T19

• [x] 10. Rewrite the commands

  • commands/{work-on,work-status,execute-tasks,finish-tasks,verify-work,create-design, create-testing-plan,contribute-to,init}.md; upgrade-the-loop.md gains the "no longer read, never deleted" note (D5)
  • Depends on: 2
  • Requirements: R1.3, R1.5, R5.1, R5.2
  • Test: T18, T19

• [x] 11. Update the docs site and capability docs

  • docs/capabilities/{spec-workflow,process-graph,review-loop,capability-docs, documentation,token-economy,testing-and-contracts}.md — requirement lines updated, one history row each
  • docs/{guide,reference,architecture,config}, docs/specs/index.md, docs/.vitepress/config.mts, README.md, CLAUDE.md, .the-loop/harness-config.schema.json's onMissing description
  • Depends on: 2
  • Requirements: R1.5, R6.2
  • Test: T18, T19

• [x] 12. Record the decision

  • docs/decisions/decision-126.md + a row in decisions.md: the log retired, the proofs rehoused, the rejected alternative (D6)
  • Depends on: none
  • Requirements: R6.1
  • Test: T18

• [x] 13. Verify and present evidence

  • Run the matrix; record rows in evidence/verification.md, the security round in evidence/security-review.md, the review rounds, docs and PRs in their own files
  • Depends on: 1–12
  • Requirements: all
  • Test: T18 (make check)

Review round 1 — the owner's review of PR #366

• [x] 14. Rename the state file and the concept

  • graph-state.json → work-item-state.json, GraphState → WorkItemState, the lock with them; the pre-rename name is read and never written, and the inner-loop scan globs both
  • Depends on: 13
  • Requirements: R7.1, R7.2, R7.3
  • Test: T20 (red→green)

• [x] 15. Move repos off artifacts — first attempt, reverted

  • A repository section on the phase-selection checklist. Rejected in review round 2: that gate runs before requirements, design and tasks exist, so it asks at the one moment nobody can answer
  • Depends on: 14
  • Requirements: R3.1
  • Test: superseded by task 17

• [x] 17. the-loop graph repos — the agent declares, once the DAG exists

  • declare_repos in the runtime (shape boundary + the instance's repositories, full set not an append, all-or-nothing); core.graphs.repos; the OpenAPI route authored first, then POST /api/v1/graph/repos; the CLI verb routed through the service; graph.repos_declared in the event catalogue; repositories reaches Runtime.config
  • The agent is told: the skill's multi-repo rule, reference/workflow.md, create-tasks-plan (the moment) and execute-tasks (the fallback), plus the CLI page
  • Depends on: 15
  • Requirements: R3.1, R3.2, R3.3, R7.4, R7.5
  • Test: T21, T22, T23 (red→green)

• [x] 16. Record the decision and re-document both

  • docs/decisions/decision-127.md + index row, rewritten after round 2 rather than amended — its reasoning was what was wrong; decision-126's superseded bullet cross-linked; capability history rows; the skill, commands, site and .gitignore
  • Depends on: 14, 17
  • Requirements: R6.1, R6.2
  • Test: T18, T19

Execution DAG

flowchart LR
    T1["1 templates"] --> T2["2 graphs"]
    T2 --> T3["3 delete log-entry"]
    T1 --> T6["6 manifest"]
    T2 --> T6
    T4["4 repos: → tasks.md"] --> T7["7 tests"]
    T5["5 seed artifacts"] --> T7
    T3 --> T7
    T6 --> T7
    T6 --> T8["8 e2e fixtures"]
    T2 --> T9["9 skill"]
    T2 --> T10["10 commands"]
    T2 --> T11["11 docs"]
    T12["12 decision"] --> T13["13 verify"]
    T7 --> T13
    T8 --> T13
    T9 --> T13
    T10 --> T13
    T11 --> T13