Anandan 739a47e3ea perf(agent): preserve prompt cache prefixes (#760) ...

* perf(agent): preserve prompt cache prefixes

Build the system prompt once per run, fingerprint the exact emitted prompt and ordered tool definitions, and expose the full system-prompt hash in traces.

Add normalized and serialized two-turn regressions covering append-only messages, stable tools, and a stable session cache key.

* fix(tools): stabilize registry iteration order

Return registered tools in name order so provider-visible tool lists and prompt-prefix fingerprints cannot drift with Go map iteration. Add a regression covering the registry ordering contract.

2026-07-19 20:29:58 +05:30

..

apply_patch_paths_test.go

Audit tail: remaining medium/low correctness fixes (CLI, config, OAuth, tools, LSP, TUI) (#275)

2026-06-20 22:21:48 +05:30

apply_patch.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

args.go

fix(sandbox): retry approved network shell commands (#301)

2026-06-23 17:44:46 +05:30

argtolerance_test.go

Tools: arg-tolerance, ask_user, secret scrubbing, structured results, hardening (#119)

2026-06-07 17:13:27 +05:30

argtolerance.go

Tools: arg-tolerance, ask_user, secret scrubbing, structured results, hardening (#119)

2026-06-07 17:13:27 +05:30

ask_user_format_test.go

polish v2: smoothness & delight (#484)

2026-07-05 13:15:14 +05:30

ask_user_test.go

feat(tui): ask_user suggested answers with a recommended default (#326)

2026-06-27 14:12:41 +05:30

ask_user.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

bash_auto_allow_test.go

fix(sandbox): retry approved network shell commands (#301)

2026-06-23 17:44:46 +05:30

bash_budget_test.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

bash_proc_unix.go

fix(tools): fix cmd.exe /S/C corrupting commands with embedded quotes (#465)

2026-07-05 11:36:00 +08:00

bash_proc_windows.go

fix(windows): resolve absolute path for taskkill to prevent hijacking (#617)

2026-07-10 10:08:11 +08:00

bash_secrets_test.go

Tool quality: per-tool denial categorization + secret scanning (#137)

2026-06-08 09:08:51 +05:30

bash_tool_test.go

fix(tools): Windows cmd.exe quoting guidance and clipboard escaping fix (#468)

2026-07-09 09:23:40 +08:00

bash.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

builtin_catalog.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

capabilities_test.go

perf(agent): concurrent read-only tool batches via capability gate (#715)

2026-07-17 17:52:11 +05:30

capabilities.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

deferred_test.go

Move deferred tool discovery into tool search (#324)

2026-06-27 13:02:09 +05:30

deferred.go

Move deferred tool discovery into tool search (#324)

2026-06-27 13:02:09 +05:30

diff_preview.go

feat(tools): show red/green diff preview on write_file and edit_file (#318)

2026-06-25 19:24:27 +05:30

edit_file.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

edit_replacers_test.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

edit_replacers.go

chore: fix golangci-lint findings (dead code, ineffassign, staticcheck) (#527)

2026-07-06 08:09:32 +08:00

escalate_model_test.go

Mid-run model escalation: escalate_model tool + --allow-escalation (#146)

2026-06-08 20:50:32 +05:30

escalate_model.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

exec_command_test.go

fix(tools): preserve SysProcAttr during PTY fallback (#618)

2026-07-10 10:08:32 +08:00

exec_command.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

exec_pty_fallback.go

Add managed background terminals (#270)

2026-06-20 20:30:39 +05:30

exec_pty_linux.go

fix: resolve 15 findings from the 2026-06-20 deep audit (#281)

2026-06-21 00:44:23 +05:30

exec_transport.go

fix(tools): preserve SysProcAttr during PTY fallback (#618)

2026-07-10 10:08:32 +08:00

file_safety_test.go

Edit safety: file read/write tracking, external-modification detection, post-apply hooks (#172)

2026-06-11 20:34:35 +05:30

file_tools_test.go

perf(grep): stop content scan after head limit (#601)

2026-07-09 09:33:31 +08:00

file_tracker_test.go

fix: warn about untracked scratch files left behind after a run (#571)

2026-07-09 09:25:09 +08:00

file_tracker.go

fix: warn about untracked scratch files left behind after a run (#571)

2026-07-09 09:25:09 +08:00

format_on_write_test.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

format_on_write.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

glob.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

grep.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

inline_diagnostics_test.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

inline_diagnostics.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

line_reader.go

Stream tool file IO (#565)

2026-07-07 22:41:13 +08:00

list_directory.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

local_browser_test.go

Add local control browser and terminal tools (#319)

2026-06-26 08:22:56 +05:30

local_browser.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

local_capture_test.go

Add local control browser and terminal tools (#319)

2026-06-26 08:22:56 +05:30

local_capture.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

local_desktop_terminal_test.go

Add local control browser and terminal tools (#319)

2026-06-26 08:22:56 +05:30

local_desktop.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

local_terminal.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

lsp_navigate_test.go

feat: five capability features from the reference-agent comparison (#276)

2026-06-20 23:12:15 +05:30

lsp_navigate.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

mutation_targets_test.go

Tools: arg-tolerance, ask_user, secret scrubbing, structured results, hardening (#119)

2026-06-07 17:13:27 +05:30

mutation_targets.go

Tools: arg-tolerance, ask_user, secret scrubbing, structured results, hardening (#119)

2026-06-07 17:13:27 +05:30

output_boundary_test.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

output_boundary.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

output_budget_test.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

output_budget.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

output_ceiling_test.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

output_ceiling.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

output_policies_test.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

output_policies.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

plan_tool_test.go

fix: behavioral-audit follow-ups (exec -p, cron/version/specialist UX, update_plan coercion, Makefile) (#286)

2026-06-21 15:57:38 +05:30

preview_test.go

feat(tui): clickable plan-step detail + plan/sidebar UX & rendering polish (#315)

2026-06-25 20:23:29 +05:30

preview.go

feat(tui): clickable plan-step detail + plan/sidebar UX & rendering polish (#315)

2026-06-25 20:23:29 +05:30

read_exclusions_test.go

feat(sandbox): harden internal/sandbox — re-entrancy guard, SOCKS5 egress, sandbox-aware search, fine-grained path lists (#199)

2026-06-14 17:37:40 +05:30

read_exclusions.go

feat(sandbox): harden internal/sandbox — re-entrancy guard, SOCKS5 egress, sandbox-aware search, fine-grained path lists (#199)

2026-06-14 17:37:40 +05:30

read_file.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

read_minified_file_test.go

Fix token accounting and usage display (#304)

2026-06-24 08:18:34 +05:30

read_minified_file.go

perf(output): add token-aware semantic output budgeting (PR11) (#717)

2026-07-18 00:02:43 +05:30

registry_test.go

perf(agent): preserve prompt cache prefixes (#760)

2026-07-19 20:29:58 +05:30

registry.go

perf(agent): preserve prompt cache prefixes (#760)

2026-07-19 20:29:58 +05:30

request_permissions.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

resource_keys.go

perf(agent): concurrent read-only tool batches via capability gate (#715)

2026-07-17 17:52:11 +05:30

sandbox_denial_test.go

fix(tools): classify silent wrapped Windows command failures as sandbox denials (#659)

2026-07-14 15:13:05 +08:00

sandbox_denial.go

fix(tools): classify silent wrapped Windows command failures as sandbox denials (#659)

2026-07-14 15:13:05 +08:00

scaffold_test.go

wedge stage 10: distribution (skill/plugin install, tools make) (#184)

2026-06-13 10:43:31 +05:30

scaffold.go

wedge stage 10: distribution (skill/plugin install, tools make) (#184)

2026-06-13 10:43:31 +05:30

search_cancel.go

fix(tools): make grep and glob respect run cancellation (#464)

2026-07-05 11:35:01 +08:00

search_cancellation_test.go

perf(grep): stop content scan after head limit (#601)

2026-07-09 09:33:31 +08:00

shell_runtime_test.go

fix(tools): block MSYS and WSL shells under the Windows sandbox (#587)

2026-07-08 17:57:56 +08:00

shell_runtime.go

fix(tools): Windows cmd.exe quoting guidance and clipboard escaping fix (#468)

2026-07-09 09:23:40 +08:00

skill_test.go

Skills & plugins: skill tool, skills loader, plugin manifests (#123)

2026-06-07 21:35:11 +05:30

skill.go

perf(agent): concurrent read-only tool batches via capability gate (#715)

2026-07-17 17:52:11 +05:30

spill_hardening_test.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

spill_owner_unix.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

spill_owner_windows.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

spill_test.go

perf: universal tool-output ceiling with spill + async post-edit diagnostics (#518)

2026-07-05 23:50:31 +08:00

spill.go

perf: universal tool-output ceiling with spill + async post-edit diagnostics (#518)

2026-07-05 23:50:31 +08:00

tool_io_bench_test.go

Stream tool file IO (#565)

2026-07-07 22:41:13 +08:00

tool_search_eager_test.go

tools: tool_search redirects eager-tool lookups instead of "no tools matched" (#259)

2026-06-19 20:02:48 +05:30

tool_search_test.go

polish v2: smoothness & delight (#484)

2026-07-05 13:15:14 +05:30

tool_search.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

types_test.go

Add local control browser and terminal tools (#319)

2026-06-26 08:22:56 +05:30

types.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

update_plan.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

web_fetch_markdown_test.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

web_fetch_markdown.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

web_fetch_test.go

chore: fix golangci-lint findings (dead code, ineffassign, staticcheck) (#527)

2026-07-06 08:09:32 +08:00

web_fetch.go

perf(agent): concurrent read-only tool batches via capability gate (#715)

2026-07-17 17:52:11 +05:30

web_search_searxng_test.go

tools: defer swarm tool schemas + keyless SearXNG web_search backend (#258)

2026-06-19 20:02:26 +05:30

web_search_test.go

fix(tools): require permission before web_search requests (#382)

2026-07-02 22:14:58 +08:00

web_search.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

workspace.go

feat: agent quality, caching, retry, and tooling upgrades (#506)

2026-07-05 17:25:26 +08:00

write_file.go

perf(tools): add explicit effect metadata for safe concurrency (#705)

2026-07-17 12:48:36 +08:00

write_tools_test.go

fix: warn about untracked scratch files left behind after a run (#571)

2026-07-09 09:25:09 +08:00