Issues

fix(ai): commandcode protocol sends the server cwd path and top-level directory listing to api.commandcode.ai on every generate call — disposition the disclosure and minimize

developing-today-forks/opencode#1948 · opened

on Nov 7, 2026

2

chore(ci): disposition open PR #1778 — superseded by merged #1783 (solid-js patch regenerated, not removed); merging it would break install

developing-today-forks/opencode#1947 · opened

on Nov 7, 2026

fix(ai): commandcode threadId is silently dropped when non-UUID and permissionMode is unvalidated — both untested (#1749)

developing-today-forks/opencode#1946 · opened

on Nov 7, 2026

fix(ai): claude-code and commandcode protocol routes still carry pre-rename provider ids — the bare route exports split from the #1823 rename

developing-today-forks/opencode#1945 · opened

on Nov 7, 2026

1

test(ai): quota figure contract — figure inventory + shared figure probe matrix for the 13 quota providers (design #1942, phase 1)

developing-today-forks/opencode#1943 · opened

on Nov 7, 2026

1

design(ai): a figure that cannot exist is not a figure — one figure contract for the quota parsers' amounts, balances, counts and percents

developing-today-forks/opencode#1942 · opened

on Nov 7, 2026

fix(ai): antigravity quota publishes integer-quantized percents — 99.9% used reads as 100% / 0% remaining, 0.1% used reads as 0%

developing-today-forks/opencode#1941 · opened

on Nov 7, 2026

2

docs(ai): quota window-contract inventory verdicts do not reflect the open known-lossy filings — six rows certify or omit parsers with live lossy issues, and turso's row has no verdict at all

developing-today-forks/opencode#1939 · opened

on Nov 7, 2026

1

test(ai): quota.test.ts describes for #1744/#1746/#1821 are nested inside the #1741 describe (from PR #1745)

developing-today-forks/opencode#1934 · opened

on Nov 7, 2026

test(ai): fidelity corpus still marks the moonshot prefixItems-with-items row known-lossy against closed #1574 — the post-#1647 fold has no live tracking (from PR #1723)

developing-today-forks/opencode#1933 · opened

on Nov 7, 2026

6

fix(core): custom-tool override order is lexicographic by file path, not config priority — a project tool can lose to a global tool of the same name (#1748)

developing-today-forks/opencode#1932 · opened

on Nov 7, 2026

fix(ai): meta quota portal read never uses the page's team — teamIdOf's page-URL branch is dead, so every portal read targets the default team

developing-today-forks/opencode#1931 · opened

on Nov 7, 2026

1

fix(ai): deepseek quota publishes a negative balance — the no-positive fallback bypasses its own >0 filter, so a negative-only payload returns "$-5.00" with ok:true

developing-today-forks/opencode#1928 · opened

on Nov 7, 2026

3

docs(ci): ci-policy.md still describes a push-to-main tier after PR #1635 removed every push trigger — check/nix-eval are dispatch-only, test is dispatch+nightly

developing-today-forks/opencode#1927 · opened

on Nov 7, 2026

5

fix(ai): antigravity convertSchema merges a multi-valued type union into the node's own anyOf/oneOf — conjunction becomes disjunction (constraints dropped on Gemini, valid calls rejected on Anthropic)

developing-today-forks/opencode#1926 · opened

on Nov 7, 2026

7

fix(ai): codex quota credits publishes a negative balance — "-5" as the credits_balance label, and a negative-only payload still returns ok:true

developing-today-forks/opencode#1924 · opened

on Nov 7, 2026

6

fix(ai): quota listing ignores disabled_providers for turso — the built-in fetcher is fetched on every default run even when disabled

developing-today-forks/opencode#1923 · opened

on Nov 7, 2026

7

fix(ai): claude quota extra-usage spend publishes negative amounts — "$-5.00 / $10.00", and a negative-only payload still returns ok:true

developing-today-forks/opencode#1919 · opened

on Nov 7, 2026

2

Sessions stall in a loop of ~4-token no-message completions (claude-sonnet-5-5 via claude-code provider)

developing-today-forks/opencode#1903 · opened

on Nov 7, 2026

3

test(core): tool-schema meta-validation inventory is hand-maintained with no completeness check — a new built-in tool escapes #1668 validation silently (#1674 follow-up)

developing-today-forks/opencode#1902 · opened

on Nov 7, 2026

4

fix(ai): muse-direct minted-key cache is process-global and ignores the credential source — a second source silently spends the first account's key

developing-today-forks/opencode#1899 · opened

on Nov 7, 2026

5

fix(function): githubActionsJWKSet cache key ignores jose options — a later caller with different options silently gets the first caller's set

developing-today-forks/opencode#1898 · opened

on Nov 6, 2026

5

fix(ai): groq quota accepts remaining above the limit — "150 / 100 requests remaining" publishes beside 0% used / 100% remaining

developing-today-forks/opencode#1897 · opened

on Nov 6, 2026

10

docs: `question` tool description invites use with no detached-context warning

developing-today-forks/opencode#1890 · opened

on Nov 6, 2026

11

prompts: model-family prompts contradict each other on subagent spawning — and DeepSeek matches no family at all

developing-today-forks/opencode#1888 · opened

on Nov 6, 2026

8

Navigation Menu

Repository navigation

Issues

fix(ai): commandcode protocol sends the server cwd path and top-level directory listing to api.commandcode.ai on every generate call — disposition the disclosure and minimize

chore(ci): disposition open PR #1778 — superseded by merged #1783 (solid-js patch regenerated, not removed); merging it would break install

fix(ai): commandcode threadId is silently dropped when non-UUID and permissionMode is unvalidated — both untested (#1749)

fix(ai): claude-code and commandcode protocol routes still carry pre-rename provider ids — the bare route exports split from the #1823 rename

test(ai): quota figure contract — figure inventory + shared figure probe matrix for the 13 quota providers (design #1942, phase 1)

design(ai): a figure that cannot exist is not a figure — one figure contract for the quota parsers' amounts, balances, counts and percents

fix(ai): antigravity quota publishes integer-quantized percents — 99.9% used reads as 100% / 0% remaining, 0.1% used reads as 0%

docs(ai): quota window-contract inventory verdicts do not reflect the open known-lossy filings — six rows certify or omit parsers with live lossy issues, and turso's row has no verdict at all

test(ai): quota.test.ts describes for #1744/#1746/#1821 are nested inside the #1741 describe (from PR #1745)

test(ai): fidelity corpus still marks the moonshot prefixItems-with-items row known-lossy against closed #1574 — the post-#1647 fold has no live tracking (from PR #1723)

fix(core): custom-tool override order is lexicographic by file path, not config priority — a project tool can lose to a global tool of the same name (#1748)

fix(ai): meta quota portal read never uses the page's team — teamIdOf's page-URL branch is dead, so every portal read targets the default team

fix(ai): deepseek quota publishes a negative balance — the no-positive fallback bypasses its own >0 filter, so a negative-only payload returns "$-5.00" with ok:true

docs(ci): ci-policy.md still describes a push-to-main tier after PR #1635 removed every push trigger — check/nix-eval are dispatch-only, test is dispatch+nightly

fix(ai): antigravity convertSchema merges a multi-valued type union into the node's own anyOf/oneOf — conjunction becomes disjunction (constraints dropped on Gemini, valid calls rejected on Anthropic)

fix(ai): codex quota credits publishes a negative balance — "-5" as the credits_balance label, and a negative-only payload still returns ok:true

fix(ai): quota listing ignores disabled_providers for turso — the built-in fetcher is fetched on every default run even when disabled

fix(ai): claude quota extra-usage spend publishes negative amounts — "$-5.00 / $10.00", and a negative-only payload still returns ok:true

Sessions stall in a loop of ~4-token no-message completions (claude-sonnet-5-5 via claude-code provider)

test(core): tool-schema meta-validation inventory is hand-maintained with no completeness check — a new built-in tool escapes #1668 validation silently (#1674 follow-up)

fix(ai): muse-direct minted-key cache is process-global and ignores the credential source — a second source silently spends the first account's key

fix(function): githubActionsJWKSet cache key ignores jose options — a later caller with different options silently gets the first caller's set

fix(ai): groq quota accepts remaining above the limit — "150 / 100 requests remaining" publishes beside 0% used / 100% remaining

docs: `question` tool description invites use with no detached-context warning

prompts: model-family prompts contradict each other on subagent spawning — and DeepSeek matches no family at all