You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Four SKILL.md hubs exceed the 5,000-token post-compaction re-attach slice, dropping their own gates #4255
Per https://code.claude.com/docs/en/skills.md (fetched 2026-09-19, 115,336 bytes, line 518): "Claude Code re-attaches the most recent invocation of each skill after the summary, keeping the first 5,000 tokens of each. Re-attached skills share a combined budget of 25,000 tokens." Line 308 adds "so every line is a recurring token cost". Line 471's 500-line tip is a separate budget these files satisfy.
Measured on disk, tier E1; the bytes-to-tokens conversion below is judgment.
discovery/0.19.18/skills/research/SKILL.md: 35,130 bytes, 258 lines, 5,212 words, over a 5,000-token window unconverted
implementation/0.16.8/skills/implement-dispatch/SKILL.md: 25,187 bytes, 126 lines, no reference/ offload; cross-references five sections of the 28,906-byte implement/SKILL.md, so the orchestrator holds both
The planning-plan/20260919T205610Z packet (F7) reproduced the cut: at roughly 4 characters per token, 5,000 tokens is about 20,000 bytes, and head -c 20000 SKILL.md | wc -l is 148 lines, ending inside Step 3. Step 4.7's outcome gate and Step 5, the approval gate the skill calls "the point", fall outside the slice. Its evidence-2.md (D) records an unresolved discrepancy: the session log says about 25 KB loaded against 42,249 bytes on disk, so the boundary is approximate but holds at either figure.
explore is inverted: SKILL.md:35 states "keeping that out of the orchestrator's context window is the point", yet lines 86-235 are worker content the worker gets again via its own skills: preload. contract.test.sh:217-222 (assertion 8) pins the equivalent fix for research; nothing pins explore.
Cheapest fix per hub: delete spoke-duplicated content, leave a load-when pointer. In plan, eight of nine Step 4.5 substeps and the Step 4.6 gate body already sit in context/plan-template.md and context/tag-decisions.md, roughly 4 KB. Two hub-only clauses must move, not be deleted: "hard-to-reverse decisions escalate EARLY" at SKILL.md:212 and the agent-teams routing surface at :199/:207. In explore, move lines 86-235 into reference/workflow.md. Fuller fix: reorder each hub so mandatory gates sit inside the first ~20 KB.
Verification: extend contract.test.sh assertion 8 to each hub against its largest spoke; head -c 20000 SKILL.md must contain every gate the skill treats as mandatory.
Comment only; this issue stays closed. This comes from the 2026-09-29 audit of the Cursor agent run. Status per hub, with sizes measured on the fix branch (PR #5301) by head -c 20000:
explore: explore/SKILL.md is 23,226 bytes, with the outcome and acceptance gates pinned in the slice (contract.test.shassert_in_slice).
planning, implementation, source-control: covered by their own reattach-slice tests (plugins/planning/tests/reattach-slice.test.sh, plugins/implementation/skills/implement-dispatch/reattach-slice.test.sh, plugins/source-control/skills/worktree/reattach-slice.test.sh). These belong to other fix groups and were not verified here.
So not all five hubs are confirmed done. #5301 uses Refs #4255.
Activity
fix(skills): keep hub gates inside the compaction re-attach slice
chore(changelog): move the #4255 source-control hold to 0.62.11
docs(implementation): keep the phase-verifier rule in the compaction …
fix(discovery): repoint blindspot and researcher at the spokes and ad…
kyle-sexton commentedon Sep 29, 2026
Comment only; this issue stays closed. This comes from the 2026-09-29 audit of the Cursor agent run. Status per hub, with sizes measured on the fix branch (PR #5301) by
head -c 20000:explore/SKILL.mdis 23,226 bytes, with the outcome and acceptance gates pinned in the slice (contract.test.shassert_in_slice).research/SKILL.mdnow keeps the Routing gate, Topic slot, Disciplines, Effort ceiling and the outcome gate with its Owner column inside the slice, with newassert_in_slicerows. fix(discovery): close the audit findings in research, agents and the parent contract #5301 closes the gap PR fix(skills): keep hub gates inside the compaction re-attach slice (#4255) #5083 left.plugins/planning/tests/reattach-slice.test.sh,plugins/implementation/skills/implement-dispatch/reattach-slice.test.sh,plugins/source-control/skills/worktree/reattach-slice.test.sh). These belong to other fix groups and were not verified here.So not all five hubs are confirmed done. #5301 uses
Refs #4255.fix(planning): keep prd and design hub gates inside the re-attach slice
fix(discovery): close the audit findings in research, agents and the …
docs(implementation): audit fixes for dispatch dedupe, verifier recor…
fix(planning): audit remediation (interview scope, outcome gate, plan…
docs(implementation): verify only behavior-changing phases in impleme…