Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
18 commits
Select commit Hold shift + click to select a range
f2d4bc3
fix: keep jCodeMunch route execution off in four carriers
Oct 3, 2026
e0c6b38
chore: register jCodeMunch route correction evidence
Oct 3, 2026
ba2f89f
docs: state the measured base and observed results in the jCodeMunch …
Oct 3, 2026
f95d4e3
chore: re-register the jCodeMunch addendum after the review repair
Oct 3, 2026
291a685
Hot-file protocol: reset manifests/evidence.json to the merge base be…
Oct 4, 2026
10e6e1f
Register this branch's files on the merge-base registry (hot-file pro…
Oct 4, 2026
7ec6ba3
Merge origin/main (1a64e8f3) into the jCodeMunch route-carrier port
Oct 4, 2026
87764f8
Hot-file protocol: reset manifests/evidence.json to main (1a64e8f3) b…
Oct 4, 2026
a529f72
Register this branch's files on main's registry (hot-file protocol: e…
Oct 4, 2026
8ca461a
Merge main 11967c851 for the custody merge (hot-file protocol: main's…
Oct 4, 2026
984b72f
Hot-file protocol: reset manifests/evidence.json to the merge base be…
Oct 4, 2026
0d15150
Register this branch's files on the merge-base registry (hot-file pro…
Oct 4, 2026
771181a
Merge main f474f6d22 for the custody merge (hot-file protocol: main's…
Oct 4, 2026
3eef1aa
Hot-file protocol: reset manifests/evidence.json to the merge base be…
Oct 4, 2026
a44180d
Reapply this branch's manifests/evidence.json edits on the merge base…
Oct 4, 2026
f5697ce
Merge main ba0e8c48f for the custody merge (hot-file protocol: main's…
Oct 4, 2026
9b096cc
Hot-file protocol: reset manifests/evidence.json to the merge base be…
Oct 4, 2026
afee0ae
Reapply this branch's manifests/evidence.json edits on the merge base…
Oct 4, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions adoption/hooks/claude/SHA256SUMS
Original file line number Diff line number Diff line change
@@ -1,10 +1,10 @@
c0979c22a701538a267e8f6b980dd2228e79db0c0b2b6bcf8d91dd2681735fb7 currency-due-notice.py
2a5f087f2c10437bab302e19469ae6b2e715828083a8bae07a7b177540b63e0e effort-default-guard.py
7b1ec8fa5c6127060c81e23b80e5fe9f7dfec9af045960e928e38763bbb95790 ../../../scripts/hooks/secret_path_guard.py
d6c3c19d6e94d3f48d2788963c8c9be45ec57e6bb81cab92b2c30d9fd5c8e8d6 token-lanes-block.md
c81a91c4ffb2fa906feb403d5a11bd294b6cf2db45795d80f9acd662c5004973 token-lanes-block.builder.md
47cc80b262468557b8cfe5c243598030757270674bc7bf286fba8c0c573abc28 token-lanes-block.researcher.md
19857a44a201948aa1d1fc9be3e24d5f3f7cc385f333ab3893411889258af4ed token-lanes-block.reviewer.md
dcd6996e9247f3c8f3d543545eb958518236b21449492ddf92721590aa3f60c8 token-lanes-block.md
2f7a53031e6eabf6b9d08989b619b74204fa36dac2559e520fb72b06225a81d3 token-lanes-block.builder.md
149ac69352a788ad67c84b4d545cab461015c4e73a9ba4c8675259402c955e4a token-lanes-block.researcher.md
ec7c6b2a80c58e0156daab25b7cfd4eea6f2e3eac77137067051601045f9d2f8 token-lanes-block.reviewer.md
81784c2edc0cc95c0c0eb3b443b5b228dd41e970236239553c66af90ddfc7eae token-lanes-block.scout.md
a2eb3298126efd82622150e5717f10eb700f4aa3076d027c86ae1fe653f29eb0 token-lanes-block.verifier.md
81e28a7c2f5429301d3306dc5ee0ecf809c67ccded62a43a98281baa8f92fedd token-lanes-subagent-start.py
2 changes: 1 addition & 1 deletion adoption/hooks/claude/token-lanes-block.builder.md
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@ TOKEN LANES (source: docs/token-session-handbook.md, "Token lanes carried into s
- Use ctx_search with specific queries and limit <=3.
- For output over ~5 KB, use ctx_batch_execute or ctx_execute with intent (indexes output; returns only titles/previews), then ctx_search; print derived answers, keep failures and original-output recovery. Pass ctx_execute cwd = the owned worktree your brief names for every language; without it, non-shell code runs at the server's project root (the coordinator's checkout) and its writes persist. Rust runs in temp; use absolute project paths. Open authorized scratch files from Python inside ctx_execute; ctx_execute_file enforces project boundaries (#852). Never bypass permissions.
- Automatic RTK: a Bash call containing $(...), backticks, <(...), a redirect to a file or a heredoc is never rewritten. && chains and multi-line blocks are rewritten segment by segment. In a pipeline RTK rewrites a supported producer feeding cat, head or tail (git diff | head -n 5) or a final grep/rg stage, but not gh (gh pr view 1 | head -n 5 is not rewritten). Run heavy reads as their own call. Never directly prefix rtk before `git show REV:path`, `diff`, `git branch`, `git log`, `jq`, `cd`, or `find` on a possibly-missing path. Use native commands or `rtk proxy <cmd>` for exact bytes/exit status; process large results in ctx_execute.
- Use Serena find_symbol / find_referencing_symbols for exact symbols and references; jcodemunch route(task, repo?, execute?), order(action, args) on indexed repos; socraticode codebase_search(query, projectPath) with explicit projectPath for conceptual questions. Open original source before judging or editing.
- Use Serena find_symbol / find_referencing_symbols for exact symbols and references; jcodemunch route(task, repo: ".") (no execute), order(action, own args) on indexed repos; socraticode codebase_search(query, projectPath) with explicit projectPath for conceptual questions. Open original source before judging or editing.
- Use ai-memory memory_query with workspace/project from .ai-memory.toml as historical evidence only, never authority.
- Use TOON for uniform arrays of flat records (same keys in every item); keep compact JSON for nested or non-uniform data, where TOON can be larger (upstream README).
- Use one lane per artifact; never stack compressors or claim token savings. Agents told to return output unmodified skip output-routing and footer rules; otherwise list token tools used and why at the end of your return.
Expand Down
2 changes: 1 addition & 1 deletion adoption/hooks/claude/token-lanes-block.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@ TOKEN LANES (source: docs/token-session-handbook.md, "Token lanes carried into s
- Use ctx_search with specific queries and limit <=3.
- For output over ~5 KB, use ctx_batch_execute or ctx_execute with intent (indexes output; returns only titles/previews), then ctx_search; print derived answers, keep failures and original-output recovery. Pass ctx_execute cwd = your working directory for every language; without it, non-shell code runs at the server's project root (the coordinator's checkout) and its writes persist. Rust runs in temp; use absolute project paths. Open authorized scratch files from Python inside ctx_execute; ctx_execute_file enforces project boundaries (#852). Never bypass permissions.
- Automatic RTK: a Bash call containing $(...), backticks, <(...), a redirect to a file or a heredoc is never rewritten. && chains and multi-line blocks are rewritten segment by segment. In a pipeline RTK rewrites a supported producer feeding cat, head or tail (git diff | head -n 5) or a final grep/rg stage, but not gh (gh pr view 1 | head -n 5 is not rewritten). Run heavy reads as their own call. Never directly prefix rtk before `git show REV:path`, `diff`, `git branch`, `git log`, `jq`, `cd`, or `find` on a possibly-missing path. Use native commands or `rtk proxy <cmd>` for exact bytes/exit status; process large results in ctx_execute.
- Use Serena find_symbol / find_referencing_symbols for exact symbols and references; jcodemunch route(task, repo?, execute?), menu(query?), order(action, args) on indexed repos; socraticode codebase_search(query, projectPath) with explicit projectPath for conceptual questions. Open original source before judging or editing.
- Use Serena find_symbol / find_referencing_symbols for exact symbols and references; jcodemunch route(task, repo: ".") (no execute), menu(query?), order(action, own args) on indexed repos; socraticode codebase_search(query, projectPath) with explicit projectPath for conceptual questions. Open original source before judging or editing.
- Use codebase-memory trace_path (include_evidence=true adds resolver class and confidence; this parameter is on trace_path, not search_graph) or search_graph for symbol/reference queries; treat edges below confidence 0.5 as unverified candidates, and verify exact caller lists with Serena find_referencing_symbols.
- Use qmd query with collections (foundation-docs, foundation-adoption, us-equities-foundation, us-equities-catalog), then get a line window. Use ai-memory memory_query with workspace/project from .ai-memory.toml as historical evidence only, never authority.
- Use TOON for uniform arrays of flat records (same keys in every item); keep compact JSON for nested or non-uniform data, where TOON can be larger (upstream README).
Expand Down
2 changes: 1 addition & 1 deletion adoption/hooks/claude/token-lanes-block.researcher.md
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@ TOKEN LANES (source: docs/token-session-handbook.md, "Token lanes carried into s
- Use ctx_search with specific queries and limit <=3.
- For output over ~5 KB, use ctx_batch_execute or ctx_execute with intent (indexes output; returns only titles/previews), then ctx_search; print derived answers, keep failures and original-output recovery. Pass ctx_execute cwd = your working directory for every language; without it, non-shell code runs at the server's project root (the coordinator's checkout) and its writes persist. Rust runs in temp; use absolute project paths. Open authorized scratch files from Python inside ctx_execute; ctx_execute_file enforces project boundaries (#852). Never bypass permissions.
- Automatic RTK: a Bash call containing $(...), backticks, <(...), a redirect to a file or a heredoc is never rewritten. && chains and multi-line blocks are rewritten segment by segment. In a pipeline RTK rewrites a supported producer feeding cat, head or tail (git diff | head -n 5) or a final grep/rg stage, but not gh (gh pr view 1 | head -n 5 is not rewritten). Run heavy reads as their own call. Never directly prefix rtk before `git show REV:path`, `diff`, `git branch`, `git log`, `jq`, `cd`, or `find` on a possibly-missing path. Use native commands or `rtk proxy <cmd>` for exact bytes/exit status; process large results in ctx_execute.
- Use Serena find_symbol / find_referencing_symbols for exact symbols and references; jcodemunch route(task, repo?, execute?), menu(query?), order(action, args) on indexed repos. Open original source before judging or editing.
- Use Serena find_symbol / find_referencing_symbols for exact symbols and references; jcodemunch route(task, repo: ".") (no execute), menu(query?), order(action, own args) on indexed repos. Open original source before judging or editing.
- Use qmd query with collections (foundation-docs, foundation-adoption, us-equities-foundation, us-equities-catalog), then get a line window. Use ai-memory memory_query with workspace/project from .ai-memory.toml as historical evidence only, never authority.
- Use TOON for uniform arrays of flat records (same keys in every item); keep compact JSON for nested or non-uniform data, where TOON can be larger (upstream README).
- Use one lane per artifact; never stack compressors or claim token savings. Agents told to return output unmodified skip output-routing and footer rules; otherwise list token tools used and why at the end of your return.
2 changes: 1 addition & 1 deletion adoption/hooks/claude/token-lanes-block.reviewer.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@ TOKEN LANES (source: docs/token-session-handbook.md, "Token lanes carried into s
- Load deferred token tools in ONE ToolSearch call before first use: "select:mcp__plugin_context-mode_context-mode__ctx_execute,mcp__plugin_context-mode_context-mode__ctx_batch_execute,mcp__plugin_context-mode_context-mode__ctx_search,mcp__serena__find_symbol,mcp__serena__find_referencing_symbols"; append task-needed ids to that same call: mcp__jcodemunch__route,mcp__jcodemunch__order,mcp__socraticode__codebase_search,mcp__ai-memory__memory_query. ToolSearch returns only tools the agent is granted. Select names exposed by the active client.
- Use ctx_search with specific queries and limit <=3.
- For output over ~5 KB, use ctx_batch_execute or ctx_execute with intent (indexes output; returns only titles/previews), then ctx_search; print derived answers, keep failures and original-output recovery. Pass ctx_execute cwd = your working directory for every language; without it, non-shell code runs at the server's project root (the coordinator's checkout) and its writes persist. Rust runs in temp; use absolute project paths. Open authorized scratch files from Python inside ctx_execute; ctx_execute_file enforces project boundaries (#852). Never bypass permissions.
- Use Serena find_symbol / find_referencing_symbols for exact symbols and references; jcodemunch route(task, repo?, execute?), order(action, args) on indexed repos; socraticode codebase_search(query, projectPath) with explicit projectPath for conceptual questions. Open original source before judging or editing.
- Use Serena find_symbol / find_referencing_symbols for exact symbols and references; jcodemunch route(task, repo: ".") (no execute), order(action, own args) on indexed repos; socraticode codebase_search(query, projectPath) with explicit projectPath for conceptual questions. Open original source before judging or editing.
- Use ai-memory memory_query with workspace/project from .ai-memory.toml as historical evidence only, never authority.
- Use TOON for uniform arrays of flat records (same keys in every item); keep compact JSON for nested or non-uniform data, where TOON can be larger (upstream README).
- Use one lane per artifact; never stack compressors or claim token savings. Agents told to return output unmodified skip output-routing and footer rules; otherwise list token tools used and why at the end of your return.
2 changes: 2 additions & 0 deletions adoption/templates/project.codex.config.template.toml
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,8 @@ startup_timeout_sec = 60
# Read front door: jcodemunch-mcp 1.108.319, counter.py FRONT_DOOR and server.py:5500-5577:
# https://github.com/jgravelle/jcodemunch-mcp/blob/8f7b34abe16fb459e0bf1c04747d584216dfe32e/src/jcodemunch_mcp/counter.py
# `order` defaults allow_state_change=false but accepts true; this list is not a read-only sandbox.
# route(execute=true) sends the whole task as the query (counter.py L616-630, blob 0605443d at 8f7b34ab and v1.108.327).
# Look up a known name with order search_symbols {repo, query: NAME}, and call route without execute for an open task.
default_tools_approval_mode = "approve"
enabled_tools = ["route", "menu", "order"]

Expand Down
Loading
Loading