fix(web): seat the out-of-capacity tab on the composer's straight edge by rynfar · Pull Request #7 · pylon-code/pylon

rynfar · 2026-08-11T22:41:11Z

The "Out of capacity" tab that offers to continue a spent thread on another account renders detached from the composer, floating beside its top-right corner instead of sitting on the edge.

Cause: the tab is positioned right-3 (12px) against the composer form, but the composer frame is rounded-[22px] — a 22px corner radius. Since 12 < 22, the tab lands on the curve. Its bottom edge is deliberately square (rounded-b-none, border-b-0) so it can read as a tab attached to a straight edge; on a curve there is nothing flush to attach to, so it looks like it is hovering next to the composer.

Fix:right-6 (24px), which clears the 22px radius so the whole tab sits on the straight run of the top edge.

Added a comment recording the constraint, since the relationship between the inset and the frame radius is invisible at the call site and easy to regress.

Note ComposerStashBadge sits on the same edge at right-4 (16px), also inside the radius. It is left alone here: it is rounded-full with a complete border, so overlapping the curve does not misread the way a square-bottomed tab does. Worth revisiting if the two are ever aligned deliberately.

Web typecheck, lint, and format clean.

Not verified visually: reproducing this needs an account actually out of capacity, which I could not stage locally. The fix is from measured geometry (12px inset vs 22px radius), not a screenshot.

Reported by the developer with a screenshot showing the tab detached at the composer's top-right.

Model: Claude Opus 5 (1M context). Harness: Claude Code.

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

github-actions · 2026-08-11T22:46:21Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.4 KiB	+26 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.4 KiB	11.3 KiB	−10 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.8 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 0df9a9e · PR result: a234a82 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

Codex decoded thread snapshot: 94.6 KiB
Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

rynfar · 2026-08-11T22:41:11Z

The "Out of capacity" tab that offers to continue a spent thread on another account renders detached from the composer, floating beside its top-right corner instead of sitting on the edge.

Cause: the tab is positioned right-3 (12px) against the composer form, but the composer frame is rounded-[22px] — a 22px corner radius. Since 12 < 22, the tab lands on the curve. Its bottom edge is deliberately square (rounded-b-none, border-b-0) so it can read as a tab attached to a straight edge; on a curve there is nothing flush to attach to, so it looks like it is hovering next to the composer.

Fix:right-6 (24px), which clears the 22px radius so the whole tab sits on the straight run of the top edge.

Added a comment recording the constraint, since the relationship between the inset and the frame radius is invisible at the call site and easy to regress.

Note ComposerStashBadge sits on the same edge at right-4 (16px), also inside the radius. It is left alone here: it is rounded-full with a complete border, so overlapping the curve does not misread the way a square-bottomed tab does. Worth revisiting if the two are ever aligned deliberately.

Web typecheck, lint, and format clean.

Not verified visually: reproducing this needs an account actually out of capacity, which I could not stage locally. The fix is from measured geometry (12px inset vs 22px radius), not a screenshot.

Reported by the developer with a screenshot showing the tab detached at the composer's top-right.

Model: Claude Opus 5 (1M context). Harness: Claude Code.

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

github-actions · 2026-08-11T22:46:21Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.4 KiB	+26 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.4 KiB	11.3 KiB	−10 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.8 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 0df9a9e · PR result: a234a82 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

Codex decoded thread snapshot: 94.6 KiB
Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

rynfar · 2026-08-11T22:41:11Z

The "Out of capacity" tab that offers to continue a spent thread on another account renders detached from the composer, floating beside its top-right corner instead of sitting on the edge.

Cause: the tab is positioned right-3 (12px) against the composer form, but the composer frame is rounded-[22px] — a 22px corner radius. Since 12 < 22, the tab lands on the curve. Its bottom edge is deliberately square (rounded-b-none, border-b-0) so it can read as a tab attached to a straight edge; on a curve there is nothing flush to attach to, so it looks like it is hovering next to the composer.

Fix:right-6 (24px), which clears the 22px radius so the whole tab sits on the straight run of the top edge.

Added a comment recording the constraint, since the relationship between the inset and the frame radius is invisible at the call site and easy to regress.

Note ComposerStashBadge sits on the same edge at right-4 (16px), also inside the radius. It is left alone here: it is rounded-full with a complete border, so overlapping the curve does not misread the way a square-bottomed tab does. Worth revisiting if the two are ever aligned deliberately.

Web typecheck, lint, and format clean.

Not verified visually: reproducing this needs an account actually out of capacity, which I could not stage locally. The fix is from measured geometry (12px inset vs 22px radius), not a screenshot.

Reported by the developer with a screenshot showing the tab detached at the composer's top-right.

Model: Claude Opus 5 (1M context). Harness: Claude Code.

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

github-actions · 2026-08-11T22:46:21Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.4 KiB	+26 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.4 KiB	11.3 KiB	−10 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.8 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 0df9a9e · PR result: a234a82 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

Codex decoded thread snapshot: 94.6 KiB
Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

rynfar · 2026-08-11T22:41:11Z

The "Out of capacity" tab that offers to continue a spent thread on another account renders detached from the composer, floating beside its top-right corner instead of sitting on the edge.

Cause: the tab is positioned right-3 (12px) against the composer form, but the composer frame is rounded-[22px] — a 22px corner radius. Since 12 < 22, the tab lands on the curve. Its bottom edge is deliberately square (rounded-b-none, border-b-0) so it can read as a tab attached to a straight edge; on a curve there is nothing flush to attach to, so it looks like it is hovering next to the composer.

Fix:right-6 (24px), which clears the 22px radius so the whole tab sits on the straight run of the top edge.

Added a comment recording the constraint, since the relationship between the inset and the frame radius is invisible at the call site and easy to regress.

Note ComposerStashBadge sits on the same edge at right-4 (16px), also inside the radius. It is left alone here: it is rounded-full with a complete border, so overlapping the curve does not misread the way a square-bottomed tab does. Worth revisiting if the two are ever aligned deliberately.

Web typecheck, lint, and format clean.

Not verified visually: reproducing this needs an account actually out of capacity, which I could not stage locally. The fix is from measured geometry (12px inset vs 22px radius), not a screenshot.

Reported by the developer with a screenshot showing the tab detached at the composer's top-right.

Model: Claude Opus 5 (1M context). Harness: Claude Code.

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

github-actions · 2026-08-11T22:46:21Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.4 KiB	+26 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.4 KiB	11.3 KiB	−10 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.8 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 0df9a9e · PR result: a234a82 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

Codex decoded thread snapshot: 94.6 KiB
Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

rynfar · 2026-08-11T22:41:11Z

The "Out of capacity" tab that offers to continue a spent thread on another account renders detached from the composer, floating beside its top-right corner instead of sitting on the edge.

Cause: the tab is positioned right-3 (12px) against the composer form, but the composer frame is rounded-[22px] — a 22px corner radius. Since 12 < 22, the tab lands on the curve. Its bottom edge is deliberately square (rounded-b-none, border-b-0) so it can read as a tab attached to a straight edge; on a curve there is nothing flush to attach to, so it looks like it is hovering next to the composer.

Fix:right-6 (24px), which clears the 22px radius so the whole tab sits on the straight run of the top edge.

Added a comment recording the constraint, since the relationship between the inset and the frame radius is invisible at the call site and easy to regress.

Note ComposerStashBadge sits on the same edge at right-4 (16px), also inside the radius. It is left alone here: it is rounded-full with a complete border, so overlapping the curve does not misread the way a square-bottomed tab does. Worth revisiting if the two are ever aligned deliberately.

Web typecheck, lint, and format clean.

Not verified visually: reproducing this needs an account actually out of capacity, which I could not stage locally. The fix is from measured geometry (12px inset vs 22px radius), not a screenshot.

Reported by the developer with a screenshot showing the tab detached at the composer's top-right.

Model: Claude Opus 5 (1M context). Harness: Claude Code.

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

github-actions · 2026-08-11T22:46:21Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.4 KiB	+26 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.4 KiB	11.3 KiB	−10 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.8 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 0df9a9e · PR result: a234a82 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

Codex decoded thread snapshot: 94.6 KiB
Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

rynfar · 2026-08-11T22:41:11Z

The "Out of capacity" tab that offers to continue a spent thread on another account renders detached from the composer, floating beside its top-right corner instead of sitting on the edge.

Cause: the tab is positioned right-3 (12px) against the composer form, but the composer frame is rounded-[22px] — a 22px corner radius. Since 12 < 22, the tab lands on the curve. Its bottom edge is deliberately square (rounded-b-none, border-b-0) so it can read as a tab attached to a straight edge; on a curve there is nothing flush to attach to, so it looks like it is hovering next to the composer.

Fix:right-6 (24px), which clears the 22px radius so the whole tab sits on the straight run of the top edge.

Added a comment recording the constraint, since the relationship between the inset and the frame radius is invisible at the call site and easy to regress.

Note ComposerStashBadge sits on the same edge at right-4 (16px), also inside the radius. It is left alone here: it is rounded-full with a complete border, so overlapping the curve does not misread the way a square-bottomed tab does. Worth revisiting if the two are ever aligned deliberately.

Web typecheck, lint, and format clean.

Not verified visually: reproducing this needs an account actually out of capacity, which I could not stage locally. The fix is from measured geometry (12px inset vs 22px radius), not a screenshot.

Reported by the developer with a screenshot showing the tab detached at the composer's top-right.

Model: Claude Opus 5 (1M context). Harness: Claude Code.

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

github-actions · 2026-08-11T22:46:21Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.4 KiB	+26 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.4 KiB	11.3 KiB	−10 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.8 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 0df9a9e · PR result: a234a82 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

Codex decoded thread snapshot: 94.6 KiB
Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

rynfar · 2026-08-11T22:41:11Z

The "Out of capacity" tab that offers to continue a spent thread on another account renders detached from the composer, floating beside its top-right corner instead of sitting on the edge.

Cause: the tab is positioned right-3 (12px) against the composer form, but the composer frame is rounded-[22px] — a 22px corner radius. Since 12 < 22, the tab lands on the curve. Its bottom edge is deliberately square (rounded-b-none, border-b-0) so it can read as a tab attached to a straight edge; on a curve there is nothing flush to attach to, so it looks like it is hovering next to the composer.

Fix:right-6 (24px), which clears the 22px radius so the whole tab sits on the straight run of the top edge.

Added a comment recording the constraint, since the relationship between the inset and the frame radius is invisible at the call site and easy to regress.

Note ComposerStashBadge sits on the same edge at right-4 (16px), also inside the radius. It is left alone here: it is rounded-full with a complete border, so overlapping the curve does not misread the way a square-bottomed tab does. Worth revisiting if the two are ever aligned deliberately.

Web typecheck, lint, and format clean.

Not verified visually: reproducing this needs an account actually out of capacity, which I could not stage locally. The fix is from measured geometry (12px inset vs 22px radius), not a screenshot.

Reported by the developer with a screenshot showing the tab detached at the composer's top-right.

Model: Claude Opus 5 (1M context). Harness: Claude Code.

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

github-actions · 2026-08-11T22:46:21Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.4 KiB	+26 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.4 KiB	11.3 KiB	−10 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.8 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 0df9a9e · PR result: a234a82 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

Codex decoded thread snapshot: 94.6 KiB
Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

rynfar · 2026-08-11T22:41:11Z

The "Out of capacity" tab that offers to continue a spent thread on another account renders detached from the composer, floating beside its top-right corner instead of sitting on the edge.

Cause: the tab is positioned right-3 (12px) against the composer form, but the composer frame is rounded-[22px] — a 22px corner radius. Since 12 < 22, the tab lands on the curve. Its bottom edge is deliberately square (rounded-b-none, border-b-0) so it can read as a tab attached to a straight edge; on a curve there is nothing flush to attach to, so it looks like it is hovering next to the composer.

Fix:right-6 (24px), which clears the 22px radius so the whole tab sits on the straight run of the top edge.

Added a comment recording the constraint, since the relationship between the inset and the frame radius is invisible at the call site and easy to regress.

Note ComposerStashBadge sits on the same edge at right-4 (16px), also inside the radius. It is left alone here: it is rounded-full with a complete border, so overlapping the curve does not misread the way a square-bottomed tab does. Worth revisiting if the two are ever aligned deliberately.

Web typecheck, lint, and format clean.

Not verified visually: reproducing this needs an account actually out of capacity, which I could not stage locally. The fix is from measured geometry (12px inset vs 22px radius), not a screenshot.

Reported by the developer with a screenshot showing the tab detached at the composer's top-right.

Model: Claude Opus 5 (1M context). Harness: Claude Code.

^{Need help on this PR? Tag @codesmith with what you need. Autofix is disabled.}

github-actions · 2026-08-11T22:46:21Z

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	11.3 KiB	11.4 KiB	+26 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	−1 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.4 KiB	11.3 KiB	−10 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+3 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.9 KiB	5.8 KiB	−13 B (−0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 0df9a9e · PR result: a234a82 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

Codex decoded thread snapshot: 94.6 KiB
Claude decoded thread snapshot: 95.4 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

fix(web): seat the out-of-capacity tab on the composer's straight edge
a234a82

github-actionsBot added vouch:trusted PR author is trusted by repo permissions or the VOUCHED list. size:XS labels Aug 11, 2026

rynfar merged commit b1df425 into pylonAug 11, 2026
11 checks passed

rynfar deleted the fix/2026-08-12-handoff-tab-position branch August 11, 2026 22:50

rynfar mentioned this pull request Aug 11, 2026
docs: record the 2026-08-12 upstream batch and advance the cursor #8
Merged

rynfar mentioned this pull request Aug 30, 2026
Prime provider parity: tasks, automation, and upstream watch #114
Open
23 tasks