2026-06-25 - 2026-07-25
Overview
22 Issues closed from 2 users
Closed
#36 fieldtest: bundle friction/spec-gap findings into collective issues instead of one-per-finding
Closed
#27 Remove the postmortem skill
Closed
#32 implement-loop resume ignores plan edits (cached extractor) and trips the discard guard after a subset commit
Closed
#25 implement-loop: anchor the quality reviewer to the loop's repo root (worktree sessions review the wrong tree)
Closed
#28 Aggregate Workflow run logs in postmortem.py — workflow sessions read as zero subagent spend
Closed
#29 Cut fixed per-iteration dispatch overhead in implement-loop; add the missing standard-mode suite gate
Closed
#30 Tier the quality review by independently-measured diff size
Closed
#31 Tier the standing reading for schema-bound workflow stages
Closed
#26 Verify agent tracker actions run under the dedicated claude account
Closed
#23 implement-loop: a task's git checkout -- <file> recovery silently discards earlier tasks' uncommitted edits to that file
Closed
#24 Named-workflow args arrive as a string: carriers silently undefined, misreported as substantive verdicts
Closed
#22 implement-loop: guard the plan-extract against silent truncation on big plans
Closed
#21 boss: after the primary task, continue to autonomously-doable open follow-ups instead of stopping at a manufactured done-state
Closed
#19 Reform brainstorm: replace the serial interview with a grounded synthetic-user swarm
Closed
#16 Define the /boss session anchor as the captured start HEAD, not the live main ref
Closed
#17 Add install.sh --force to repoint symlinks and document the worktree caveat
Closed
#18 Document the worktree slug caveat for the postmortem default log dir
Closed
#15 Add a repo CLAUDE.md forbidding worktree development of the plugin
Closed
#12 implement-loop: iteration DONE has no positive-evidence precondition (the #11 no-op-reads-as-success pattern in the main path)
Closed
#13 grounding-check: a vacuous PASS (zero assumptions extracted) auto-signs the spec under /boss
Closed
#14 Convert prose-only no-op guards to structural floors (fieldtest examples, audit drift-review)
Closed
#11 compiler-driven-edit returns DONE on a no-op edit (verdict vacuously true on a clean tree)
25 Issues created by 1 user
Opened
#11 compiler-driven-edit returns DONE on a no-op edit (verdict vacuously true on a clean tree)
Opened
#12 implement-loop: iteration DONE has no positive-evidence precondition (the #11 no-op-reads-as-success pattern in the main path)
Opened
#13 grounding-check: a vacuous PASS (zero assumptions extracted) auto-signs the spec under /boss
Opened
#14 Convert prose-only no-op guards to structural floors (fieldtest examples, audit drift-review)
Opened
#15 Add a repo CLAUDE.md forbidding worktree development of the plugin
Opened
#16 Define the /boss session anchor as the captured start HEAD, not the live main ref
Opened
#17 Add install.sh --force to repoint symlinks and document the worktree caveat
Opened
#18 Document the worktree slug caveat for the postmortem default log dir
Opened
#19 Reform brainstorm: replace the serial interview with a grounded synthetic-user swarm
Opened
#21 boss: after the primary task, continue to autonomously-doable open follow-ups instead of stopping at a manufactured done-state
Opened
#22 implement-loop: guard the plan-extract against silent truncation on big plans
Opened
#23 implement-loop: a task's git checkout -- <file> recovery silently discards earlier tasks' uncommitted edits to that file
Opened
#24 Named-workflow args arrive as a string: carriers silently undefined, misreported as substantive verdicts
Opened
#25 implement-loop: anchor the quality reviewer to the loop's repo root (worktree sessions review the wrong tree)
Opened
#26 Verify agent tracker actions run under the dedicated claude account
Opened
#27 Remove the postmortem skill
Opened
#28 Aggregate Workflow run logs in postmortem.py — workflow sessions read as zero subagent spend
Opened
#29 Cut fixed per-iteration dispatch overhead in implement-loop; add the missing standard-mode suite gate
Opened
#30 Tier the quality review by independently-measured diff size
Opened
#31 Tier the standing reading for schema-bound workflow stages
Opened
#32 implement-loop resume ignores plan edits (cached extractor) and trips the discard guard after a subset commit
Opened
#33 fieldtester: committed fixture labs must not reference the engine via the run worktree's absolute path
Opened
#34 Reconcile the task_text_path carrier contract with implement-loop's inline task-text handoff
Opened
#35 Experiment: full pipeline with implement-skills disabled — quality vs token tracking
Opened
#36 fieldtest: bundle friction/spec-gap findings into collective issues instead of one-per-finding