Issues

• subagent-driven-development: "Five rounds maximum per task" lies past the 20,000-character compaction re-attach cut, and one task ran 18+ fix rounds (Claude Code)

  obra/superpowers#2481 · Filter by author dan-1492 opened on Oct 8, 2026

• Increase usage of Claude Haiku 5.5 for improved efficiency

  enhancement
  obra/superpowers#2476 · Filter by author Cioscos opened on Oct 7, 2026

• Create docs/superpowers/README.md to tell about Superpowers in the user's repo

  enhancement
  obra/superpowers#2475 · Filter by author dolmen opened on Oct 7, 2026

• requesting-code-review: verification must cover the changed files' own build config and the real deploy path

  obra/superpowers#2474 · Filter by author edcjoel-maker opened on Oct 7, 2026

• brainstorming / writing-plans: commit the spec/plan by path so pre-staged changes are not swept in

  obra/superpowers#2473 · Filter by author edcjoel-maker opened on Oct 7, 2026
  1

• subagent-driven-development: deferred "minor" findings are graded by likelihood only — add recoverability

  obra/superpowers#2471 · Filter by author edcjoel-maker opened on Oct 7, 2026

• subagent-driven-development: interactive/manual verification steps in a brief must be bounded by command flags, not by prose

  obra/superpowers#2472 · Filter by author edcjoel-maker opened on Oct 7, 2026
  1

• subagent-driven-development: plan-verbatim code passes its own tests — reviewers need adversarial probes per named risk, including the seam with unchanged code

  obra/superpowers#2470 · Filter by author edcjoel-maker opened on Oct 7, 2026

• subagent-driven-development: document a tool-only fallback for each helper script

  obra/superpowers#2469 · Filter by author dachria-cmyk opened on Oct 6, 2026

• writing-plans: a task that reads from a row-capped service should name the cap, and its fake should enforce it

  obra/superpowers#2468 · Filter by author dachria-cmyk opened on Oct 6, 2026

• executing-plans: task-done assumes every task commits — no-commit tasks record an empty range indistinguishable from a forgotten commit

  obra/superpowers#2467 · Filter by author rexescario opened on Oct 6, 2026

• Support a handoff section (template or hook) in brainstorming specs, for downstream tools

  obra/superpowers#2465 · Filter by author JRGranell opened on Oct 6, 2026

• create desktop mobile apparatus

  obra/superpowers#2462 · Filter by author joseantonioray52-create opened on Oct 5, 2026

• user@example.com

  obra/superpowers#2461 · Filter by author joseantonioray52-create opened on Oct 5, 2026

• systematic-debugging: example prints secret value; find-polluter.sh false 'clean' result and unsafe file loop

  obra/superpowers#2460 · Filter by author gptprojectmanager opened on Oct 5, 2026
  1

• task-brief: alphanumeric task IDs ("Task A1", "Task B2") are never recognised as task headings

  obra/superpowers#2459 · Filter by author rhuthik opened on Oct 5, 2026

• executing-plans: a test-writing step should end in its own run

  obra/superpowers#2458 · Filter by author sameermkhan0650 opened on Oct 5, 2026
  1

• key need made

  obra/superpowers#2456 · Filter by author joseantonioray52-create opened on Oct 4, 2026

• executing-plans and subagent-driven-development: the main/master consent rule does not defer to a declared branch policy

  obra/superpowers#2455 · Filter by author grinidx opened on Oct 3, 2026

• Prompt audit against Claude Fable 5.1: pressure language and update suppressors in using-superpowers, executing-plans, subagent-driven-development

  obra/superpowers#2454 · Filter by author radekklein92 opened on Oct 3, 2026
  1

• writing-plans Review Focus: name three edge classes that verified plans still missed (largest allowed value, failure after a side effect, rapid successive writes)

  obra/superpowers#2453 · Filter by author abat-kek opened on Oct 3, 2026

• writing-plans / test-driven-development: a test for a protective layer is evidence only once it fails when the layer is weakened

  obra/superpowers#2452 · Filter by author abat-kek opened on Oct 3, 2026

• subagent-driven-development / executing-plans: the final report should be derived from a plan-to-diff mapping, not written freely

  obra/superpowers#2451 · Filter by author abat-kek opened on Oct 3, 2026

• subagent-driven-development: a fix round that changes an interface leaves later tasks' plan text calling the old one

  obra/superpowers#2450 · Filter by author abat-kek opened on Oct 3, 2026

• systematic-debugging: check the hypothesis against the user's own data before reporting it as the cause

  obra/superpowers#2449 · Filter by author abat-kek opened on Oct 3, 2026