@mxalbert/context-mode 2.0.0

package/configs/antigravity/GEMINI.md

@@ -0,0 +1,68 @@
+# context-mode — MANDATORY routing rules
+
+context-mode MCP tools available. Rules protect context window from flooding. One unrouted command dumps 56 KB into context. Antigravity has NO hooks — these instructions are ONLY enforcement. Follow strictly.
+
+## Think in Code — MANDATORY
+
+Analyze/count/filter/compare/search/parse/transform data: **write code** via `mcp__context-mode__ctx_execute(language, code)`, `console.log()` only the answer. Do NOT read raw data into context. PROGRAM the analysis, not COMPUTE it. Pure JavaScript — Node.js built-ins only (`fs`, `path`, `child_process`). `try/catch`, handle `null`/`undefined`. One script replaces ten tool calls.
+
+## BLOCKED — do NOT use
+
+### curl / wget — FORBIDDEN
+Do NOT use `curl`/`wget` via `run_command`. Dumps raw HTTP into context.
+Use: `mcp__context-mode__ctx_fetch_and_index(url, source)` or `mcp__context-mode__ctx_execute(language: "javascript", code: "const r = await fetch(...)")`
+
+### Inline HTTP — FORBIDDEN
+No `node -e "fetch(..."`, `python -c "requests.get(..."` via `run_command`. Bypasses sandbox.
+Use: `mcp__context-mode__ctx_execute(language, code)` — only stdout enters context
+
+### Direct web fetching — FORBIDDEN
+No `read_url_content` for large pages. Raw HTML can exceed 100 KB.
+Use: `mcp__context-mode__ctx_fetch_and_index(url, source)` then `mcp__context-mode__ctx_search(queries)`
+
+## REDIRECTED — use sandbox
+
+### Shell (>20 lines output)
+`run_command` ONLY for: `git`, `mkdir`, `rm`, `mv`, `cd`, `ls`, `npm install`, `pip install`.
+Otherwise: `mcp__context-mode__ctx_batch_execute(commands, queries)` or `mcp__context-mode__ctx_execute(language: "javascript", code: "...")`. Use `language: "shell"` only when code matches the host shell.
+
+### File reading (for analysis)
+Reading to **edit** → `view_file`/`replace_file_content` correct. Reading to **analyze/explore/summarize** → `mcp__context-mode__ctx_execute_file(path, language, code)`.
+
+### Search (large results)
+Use `mcp__context-mode__ctx_execute(language: "javascript", code: "...")` in sandbox for portable filtering/counting.
+
+## Tool selection
+
+1. **GATHER**: `mcp__context-mode__ctx_batch_execute(commands, queries)` — runs all commands, auto-indexes, returns search. ONE call replaces 30+. Each command: `{label: "header", command: "..."}`.
+2. **FOLLOW-UP**: `mcp__context-mode__ctx_search(queries: ["q1", "q2", ...])` — all questions as array, ONE call.
+3. **PROCESSING**: `mcp__context-mode__ctx_execute(language, code)` | `mcp__context-mode__ctx_execute_file(path, language, code)` — sandbox, only stdout enters context.
+4. **WEB**: `mcp__context-mode__ctx_fetch_and_index(url, source)` then `mcp__context-mode__ctx_search(queries)` — raw HTML never enters context.
+5. **INDEX**: `mcp__context-mode__ctx_index(content, source)` — store in FTS5 for later search.
+
+## Parallel I/O batches
+
+For multi-URL fetches or multi-API calls, **always** include `concurrency: N` (1-8):
+
+- `mcp__context-mode__ctx_batch_execute(commands: [3+ network commands], concurrency: 5)` — gh, curl, dig, docker inspect, multi-region cloud queries
+- `mcp__context-mode__ctx_fetch_and_index(requests: [{url, source}, ...], concurrency: 5)` — multi-URL batch fetch
+
+**Use concurrency 4-8** for I/O-bound work (network calls, API queries). **Keep concurrency 1** for CPU-bound (npm test, build, lint) or commands sharing state (ports, lock files, same-repo writes).
+
+GitHub API rate-limit: cap at 4 for `gh` calls.
+
+## Output
+
+Write artifacts to FILES — never inline. Return: file path + 1-line description.
+Descriptive source labels for `search(source: "label")`.
+
+## ctx commands
+
+| Command | Action |
+|---------|--------|
+| `ctx stats` | Call `stats` MCP tool, display full output verbatim |
+| `ctx doctor` | Call `doctor` MCP tool, run returned shell command, display as checklist |
+| `ctx upgrade` | Call `upgrade` MCP tool, run returned shell command, display as checklist |
+| `ctx purge` | Call `purge` MCP tool with confirm: true. Warns before wiping knowledge base. |
+
+After /clear or /compact: knowledge base and session stats preserved. Use `ctx purge` to start fresh.

package/configs/antigravity/mcp_config.json

@@ -0,0 +1,7 @@
+{
+  "mcpServers": {
+    "context-mode": {
+      "command": "context-mode"
+    }
+  }
+}

package/configs/antigravity-cli/hooks/hooks.json

@@ -0,0 +1,37 @@
+{
+  "hooks": {
+    "PreToolUse": [
+      {
+        "matcher": "run_command|view_file|grep_search|web_fetch|read_url_content",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "context-mode hook antigravity-cli pretooluse"
+          }
+        ]
+      }
+    ],
+    "PostToolUse": [
+      {
+        "matcher": "",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "context-mode hook antigravity-cli posttooluse"
+          }
+        ]
+      }
+    ],
+    "Stop": [
+      {
+        "matcher": "",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "context-mode hook antigravity-cli stop"
+          }
+        ]
+      }
+    ]
+  }
+}

package/configs/antigravity-cli/hooks.json

@@ -0,0 +1,37 @@
+{
+  "hooks": {
+    "PreToolUse": [
+      {
+        "matcher": "run_command|view_file|grep_search|web_fetch|read_url_content",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "context-mode hook antigravity-cli pretooluse"
+          }
+        ]
+      }
+    ],
+    "PostToolUse": [
+      {
+        "matcher": "",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "context-mode hook antigravity-cli posttooluse"
+          }
+        ]
+      }
+    ],
+    "Stop": [
+      {
+        "matcher": "",
+        "hooks": [
+          {
+            "type": "command",
+            "command": "context-mode hook antigravity-cli stop"
+          }
+        ]
+      }
+    ]
+  }
+}

package/configs/antigravity-cli/mcp_config.json

@@ -0,0 +1,10 @@
+{
+  "mcpServers": {
+    "context-mode": {
+      "command": "context-mode",
+      "env": {
+        "CONTEXT_MODE_PLATFORM": "antigravity-cli"
+      }
+    }
+  }
+}

package/configs/antigravity-cli/plugin.json

@@ -0,0 +1,14 @@
+{
+  "name": "context-mode",
+  "version": "2.0.0",
+  "description": "context-mode for Antigravity CLI (agy): sandboxed code execution, FTS5 knowledge base, and session capture. Saves your context window by keeping raw bytes out of the conversation.",
+  "author": {
+    "name": "Mert Koseoğlu",
+    "url": "https://github.com/mxalbert1996"
+  },
+  "homepage": "https://github.com/mxalbert1996/context-mode#readme",
+  "license": "Elastic-2.0",
+  "skills": [
+    "./skills/context-mode"
+  ]
+}

package/configs/antigravity-cli/rules/context-mode.md

@@ -0,0 +1,77 @@
+# context-mode routing for Antigravity CLI
+
+context-mode MCP tools are installed for this agy session. Use them when the
+task analyzes, counts, filters, compares, searches, parses, transforms, fetches,
+or otherwise processes data. Raw bytes should stay in the sandbox; only the
+derived answer should enter the conversation.
+
+## Do not dump — derive (most common mistake)
+
+Do NOT use `context-mode/ctx_execute_file` or `ctx_execute` to print a whole file
+or a full method/config (e.g. `print(FILE_CONTENT)`, `cat config.yaml`, dumping a
+281-line file). That defeats the purpose: on agy the tool's stdout is saved to a
+step file that you then read back, so a full dump costs your context window TWICE
+(once in the result, once on the follow-up read). Print only the specific value,
+matches, count, or known line-range you need:
+
+- WRONG: `ctx_execute_file(path:"config.yaml", language:"python", code:"print(FILE_CONTENT)")`
+- RIGHT (value): `code:"import yaml; d=yaml.safe_load(FILE_CONTENT); print(d['active_strategy'])"`
+- RIGHT (matches): `code:"import re;[print(i+1,l) for i,l in enumerate(FILE_CONTENT.splitlines()) if 'active_strategies' in l]"`
+- RIGHT (known slice): `code:"print(chr(10).join(FILE_CONTENT.splitlines()[190:230]))"`
+
+If you truly need to read a small, exact byte range to edit it, native `Read` /
+`view_file` on that range is fine — but never dump an entire file through a hook.
+
+## Tool call surface
+
+Antigravity CLI exposes context-mode tools as `context-mode/<tool>` calls. If
+the host uses the generic MCP wrapper, call `call_mcp_tool` with:
+
+- `ServerName`: `"context-mode"`
+- `ToolName`: `"ctx_execute"`, `"ctx_execute_file"`, `"ctx_batch_execute"`,
+  `"ctx_fetch_and_index"`, `"ctx_search"`, or `"ctx_index"`
+- `Arguments`: a JSON object for that tool
+
+Do not read `~/.gemini/antigravity-cli/mcp/context-mode/*.json` to discover
+schemas. Those files are agy's cached tool schemas and reading them spends
+context. Use these argument shapes instead:
+
+- `context-mode/ctx_execute`: `{"language":"python","code":"..."}`
+- `context-mode/ctx_execute_file`:
+  `{"path":"path/to/file","language":"python","code":"..."}`
+- `context-mode/ctx_batch_execute`:
+  `{"commands":[{"label":"...","command":"..."}],"queries":["..."]}`
+- `context-mode/ctx_fetch_and_index`: `{"url":"https://...","source":"..."}`
+- `context-mode/ctx_search`: `{"queries":["q1","q2"]}`
+- `context-mode/ctx_index`: `{"path":"path/to/file-or-dir","source":"..."}`
+  or `{"content":"...","source":"..."}`
+
+## Mandatory routing
+
+- Think in code. For analyze/count/filter/compare/search/parse/transform tasks,
+  write code with `context-mode/ctx_execute` and print only the final answer.
+  Program the analysis; do not read raw data and compute mentally.
+- File read for analysis: there is no separate `ctx_read` tool.
+  `context-mode/ctx_execute_file` is the context-mode file-read surface. It
+  loads the file into `FILE_CONTENT` inside the sandbox. Print only selected
+  lines, counts, matches, summaries, or structured results. Never print
+  `FILE_CONTENT` wholesale unless the user explicitly asks for a full file dump.
+- Native `Read` / `view_file` is correct when editing requires exact bytes, or
+  when a small known range is needed. For analysis, exploration, summarization,
+  counting, filtering, or searching inside a file, use
+  `context-mode/ctx_execute_file`.
+- Use one `context-mode/ctx_batch_execute` call for multi-command repository
+  reconnaissance. One batch should replace many shell/list/search calls.
+- Use `context-mode/ctx_execute` for shell commands whose output may exceed a
+  short fixed answer. Native Bash is only for git, mkdir, rm, mv, navigation,
+  installs, or short observable output.
+- Use `context-mode/ctx_fetch_and_index` for web content, then
+  `context-mode/ctx_search` to query it. Do not dump raw HTML into the
+  conversation.
+- Use `context-mode/ctx_index` only when content should be stored and searched
+  later. For a one-off file question, prefer `context-mode/ctx_execute_file`;
+  for follow-up retrieval, index with a descriptive `source` and query with
+  `context-mode/ctx_search`.
+- Return only derived answers, concise summaries, selected snippets, or file
+  paths to written artifacts. Do not paste raw command dumps, full files, large
+  search results, cached schemas, or raw HTML into the conversation.

package/configs/antigravity-cli/skills/context-mode/SKILL.md

@@ -0,0 +1,77 @@
+---
+name: context-mode
+description: Mandatory routing rules for Antigravity CLI. Invoke when exploring a codebase, reading files for analysis, listing context-mode tools, searching, parsing, counting, comparing, summarizing, fetching web content, or running data-heavy commands.
+---
+
+# context-mode for Antigravity CLI
+
+context-mode MCP tools are available in agy as `context-mode/<tool>`.
+Use these tools directly. Do not inspect cached schema files under
+`~/.gemini/antigravity-cli/mcp/context-mode/*.json`, and do not run
+`ctx_execute` just to discover tool names or schemas.
+
+If agy uses the generic MCP wrapper, call `call_mcp_tool` with:
+
+- `ServerName`: `"context-mode"`
+- `ToolName`: one of the tool names below
+- `Arguments`: the JSON arguments for that tool
+
+## Tool List
+
+- `context-mode/ctx_execute`: run JavaScript, TypeScript, Python, shell, or
+  other code in a sandbox. Print only the final answer.
+- `context-mode/ctx_execute_file`: read one file into `FILE_CONTENT` inside the
+  sandbox and run code over it. This is the context-mode file-read surface.
+- `context-mode/ctx_batch_execute`: run multiple repository commands in one
+  batch, index large output, and answer follow-up queries.
+- `context-mode/ctx_index`: store a file, directory, or content in the local
+  FTS5 knowledge base for later search.
+- `context-mode/ctx_search`: search indexed content and captured session
+  memory. Batch related questions in `queries`.
+- `context-mode/ctx_fetch_and_index`: fetch web content, store it, then query
+  with `ctx_search`.
+- `context-mode/ctx_stats`: show context savings and current-session stats.
+- `context-mode/ctx_doctor`: diagnose context-mode runtime and hook health.
+- `context-mode/ctx_upgrade`: provide upgrade or repair guidance.
+- `context-mode/ctx_purge`: purge stored context-mode knowledge after
+  confirmation.
+- `context-mode/ctx_insight`: launch or report the Insight analytics app.
+
+When the user asks "what context-mode tools are available", answer from this
+list. Do not list `~/.gemini/antigravity-cli/mcp/context-mode` and do not read
+the JSON schema files.
+
+## File Reading
+
+There is no separate `ctx_read` tool. Use `context-mode/ctx_execute_file` for
+file reads that analyze, summarize, extract, count, filter, or compare.
+
+Native `Read` / `view_file` is only right when editing needs exact bytes in the
+conversation, or when the user explicitly asks to view a small known range.
+
+Use JavaScript by default for lightweight file analysis:
+
+```json
+{
+  "path": "README.md",
+  "language": "javascript",
+  "code": "const lines = FILE_CONTENT.split(/\\r?\\n/); console.log(lines.slice(0, 20).join('\\n'));"
+}
+```
+
+Never print `FILE_CONTENT` wholesale unless the user explicitly asks for a full
+file dump. Print selected lines, counts, matches, summaries, or structured
+results.
+
+## Codebase Exploration
+
+For repository exploration, prefer one `context-mode/ctx_batch_execute` call
+instead of many native `ListDir`, `Read`, `Grep`, or shell calls.
+
+For one-off computed answers, use `context-mode/ctx_execute`.
+For one-file analysis, use `context-mode/ctx_execute_file`.
+For durable recall across follow-up questions, use `context-mode/ctx_index`
+with a descriptive `source`, then `context-mode/ctx_search`.
+
+Return concise derived answers. Do not paste raw command dumps, full files,
+large search results, cached schemas, or raw HTML into the conversation.

package/configs/claude-code/CLAUDE.md

@@ -0,0 +1,91 @@
+# context-mode — MANDATORY routing rules
+
+context-mode MCP tools available. Rules protect context window from flooding. One unrouted command dumps 56 KB into context.
+
+## Think in Code — MANDATORY
+
+Analyze/count/filter/compare/search/parse/transform data: **write code** via `ctx_execute(language, code)`, `console.log()` only the answer. Do NOT read raw data into context. PROGRAM the analysis, not COMPUTE it. Pure JavaScript — Node.js built-ins only (`fs`, `path`, `child_process`). `try/catch`, handle `null`/`undefined`. One script replaces ten tool calls.
+
+## BLOCKED — do NOT attempt
+
+### curl / wget — BLOCKED
+Intercepted and replaced with error. Do NOT retry.
+Use: `ctx_fetch_and_index(url, source)` or `ctx_execute(language: "javascript", code: "const r = await fetch(...)")`
+
+### Inline HTTP — BLOCKED
+`fetch('http`, `requests.get(`, `requests.post(`, `http.get(`, `http.request(` — intercepted. Do NOT retry.
+Use: `ctx_execute(language, code)` — only stdout enters context
+
+### WebFetch — BLOCKED
+Use: `ctx_fetch_and_index(url, source)` then `ctx_search(queries)`
+
+## REDIRECTED — use sandbox
+
+### Bash (>20 lines output)
+Bash ONLY for: `git`, `mkdir`, `rm`, `mv`, `cd`, `ls`, `npm install`, `pip install`.
+Otherwise: `ctx_batch_execute(commands, queries)` or `ctx_execute(language: "javascript", code: "...")`. Use `language: "shell"` only when code matches the host shell.
+
+### Read (for analysis)
+Reading to **Edit** → Read correct. Reading to **analyze/explore/summarize** → `ctx_execute_file(path, language, code)`.
+
+### Grep — may flood context
+Use `ctx_execute(language: "javascript", code: "...")` in sandbox for portable filtering/counting.
+
+## Tool selection
+
+0. **MEMORY**: `ctx_search(sort: "timeline")` — after resume, check prior context before asking user.
+1. **GATHER**: `ctx_batch_execute(commands, queries)` — runs all commands, auto-indexes, returns search. ONE call replaces 30+. Each command: `{label: "header", command: "..."}`.
+2. **FOLLOW-UP**: `ctx_search(queries: ["q1", "q2", ...])` — all questions as array, ONE call (default relevance mode).
+3. **PROCESSING**: `ctx_execute(language, code)` | `ctx_execute_file(path, language, code)` — sandbox, only stdout enters context.
+4. **WEB**: `ctx_fetch_and_index(url, source)` then `ctx_search(queries)` — raw HTML never enters context.
+5. **INDEX**: `ctx_index(content, source)` — store in FTS5 for later search.
+
+## Parallel I/O batches
+
+For multi-URL fetches or multi-API calls, **always** include `concurrency: N` (1-8):
+
+- `ctx_batch_execute(commands: [3+ network commands], concurrency: 5)` — gh, curl, dig, docker inspect, multi-region cloud queries
+- `ctx_fetch_and_index(requests: [{url, source}, ...], concurrency: 5)` — multi-URL batch fetch
+
+**Use concurrency 4-8** for I/O-bound work (network calls, API queries). **Keep concurrency 1** for CPU-bound (npm test, build, lint) or commands sharing state (ports, lock files, same-repo writes).
+
+GitHub API rate-limit: cap at 4 for `gh` calls.
+
+## Subagent routing
+
+Routing block auto-injected into subagent prompts. Bash-type subagents upgraded to general-purpose. No manual instruction needed.
+
+## Output
+
+Write artifacts to FILES — never inline. Return: file path + 1-line description.
+Descriptive source labels for `ctx_search(source: "label")`.
+
+## Session Continuity
+
+Skills, roles, and decisions persist for the entire session. Do not abandon them as the conversation grows.
+
+## Memory
+
+Session history is persistent and searchable. On resume, search BEFORE asking the user:
+
+| Need | Command |
+|------|---------|
+| What were we working on? | `ctx_search(queries: ["summary"], source: "compaction", sort: "timeline")` |
+| What was the first request? | `ctx_search(queries: ["prompt"], source: "user-prompt", sort: "timeline")` |
+| What did we decide? | `ctx_search(queries: ["decision"], source: "decision", sort: "timeline")` |
+| What NOT to repeat? | `ctx_search(queries: ["rejected"], source: "rejected-approach")` |
+| What constraints exist? | `ctx_search(queries: ["constraint"], source: "constraint")` |
+
+DO NOT ask "what were we working on?" — SEARCH FIRST.
+If search returns 0 results, proceed as a fresh session.
+
+## ctx commands
+
+| Command | Action |
+|---------|--------|
+| `ctx stats` | Call `ctx_stats` MCP tool, display full output verbatim |
+| `ctx doctor` | Call `ctx_doctor` MCP tool, run returned shell command, display as checklist |
+| `ctx upgrade` | Call `ctx_upgrade` MCP tool, run returned shell command, display as checklist |
+| `ctx purge` | Call `ctx_purge` MCP tool with confirm: true. Warns before wiping knowledge base. |
+
+After /clear or /compact: knowledge base and session stats preserved. Use `ctx purge` to start fresh.

package/configs/codex/AGENTS.md

@@ -0,0 +1,99 @@
+# context-mode — MANDATORY routing rules
+
+context-mode MCP tools available. Rules protect context window from flooding. One unrouted command dumps 56 KB into context. Codex CLI hooks provide runtime enforcement when `[features].hooks = true`; these instructions remain mandatory model-side enforcement. Follow strictly.
+
+## Think in Code — MANDATORY
+
+Analyze/count/filter/compare/search/parse/transform data: **write code** via `ctx_execute(language, code)`, `console.log()` only the answer. Do NOT read raw data into context. PROGRAM the analysis, not COMPUTE it. Pure JavaScript — Node.js built-ins only (`fs`, `path`, `child_process`). `try/catch`, handle `null`/`undefined`. One script replaces ten tool calls.
+
+## BLOCKED — do NOT use
+
+### curl / wget — FORBIDDEN
+Do NOT use `curl`/`wget` in shell. Dumps raw HTTP into context.
+Use: `ctx_fetch_and_index(url, source)` or `ctx_execute(language: "javascript", code: "const r = await fetch(...)")`
+
+### Inline HTTP — FORBIDDEN
+No `node -e "fetch(..."`, `python -c "requests.get(..."`. Bypasses sandbox.
+Use: `ctx_execute(language, code)` — only stdout enters context
+
+### Direct web fetching — FORBIDDEN
+Raw HTML can exceed 100 KB.
+Use: `ctx_fetch_and_index(url, source)` then `ctx_search(queries)`
+
+## REDIRECTED — use sandbox
+
+### Shell (>20 lines output)
+Shell ONLY for: `git`, `mkdir`, `rm`, `mv`, `cd`, `ls`, `npm install`, `pip install`.
+Otherwise: `ctx_batch_execute(commands, queries)` or `ctx_execute(language: "javascript", code: "...")`. Use `language: "shell"` only when code matches the host shell.
+
+### File reading (for analysis)
+Reading to **edit** → reading correct. Reading to **analyze/explore/summarize** → `ctx_execute_file(path, language, code)`.
+
+### grep / search (large results)
+Use `ctx_execute(language: "javascript", code: "...")` in sandbox for portable filtering/counting.
+
+## Tool selection
+
+0. **MEMORY**: `ctx_search(sort: "timeline")` — after resume, check prior context before asking user.
+1. **GATHER**: `ctx_batch_execute(commands, queries)` — runs all commands, auto-indexes, returns search. ONE call replaces 30+. Each command: `{label: "header", command: "..."}`.
+2. **FOLLOW-UP**: `ctx_search(queries: ["q1", "q2", ...])` — all questions as array, ONE call (default relevance mode).
+3. **PROCESSING**: `ctx_execute(language, code)` | `ctx_execute_file(path, language, code)` — sandbox, only stdout enters context.
+4. **WEB**: `ctx_fetch_and_index(url, source)` then `ctx_search(queries)` — raw HTML never enters context.
+5. **INDEX**: `ctx_index(content, source)` — store in FTS5 for later search.
+
+## Parallel I/O batches
+
+For multi-URL fetches or multi-API calls, **always** include `concurrency: N` (1-8):
+
+- `ctx_batch_execute(commands: [3+ network commands], concurrency: 5)` — gh, curl, dig, docker inspect, multi-region cloud queries
+- `ctx_fetch_and_index(requests: [{url, source}, ...], concurrency: 5)` — multi-URL batch fetch
+
+**Use concurrency 4-8** for I/O-bound work (network calls, API queries). **Keep concurrency 1** for CPU-bound (npm test, build, lint) or commands sharing state (ports, lock files, same-repo writes).
+
+GitHub API rate-limit: cap at 4 for `gh` calls.
+
+## Output
+
+Write artifacts to FILES — never inline. Return: file path + 1-line description.
+Descriptive source labels for `ctx_search(source: "label")`.
+
+## Session Continuity
+
+Skills, roles, and decisions persist for the entire session. Do not abandon them as the conversation grows.
+
+## Memory
+
+Session history is persistent and searchable. On resume, search BEFORE asking the user:
+
+| Need | Command |
+|------|---------|
+| What were we working on? | `ctx_search(queries: ["summary"], source: "compaction", sort: "timeline")` |
+| What did we decide? | `ctx_search(queries: ["decision"], source: "decision", sort: "timeline")` |
+| What NOT to repeat? | `ctx_search(queries: ["rejected"], source: "rejected-approach")` |
+| What constraints exist? | `ctx_search(queries: ["constraint"], source: "constraint")` |
+
+Note: user-prompt history not available.
+
+DO NOT ask "what were we working on?" — SEARCH FIRST.
+If search returns 0 results, proceed as a fresh session.
+
+## ctx commands
+
+| Command | Action |
+|---------|--------|
+| `ctx stats` | Call `stats` MCP tool, display full output verbatim |
+| `ctx doctor` | Call `doctor` MCP tool, run returned shell command, display as checklist |
+| `ctx upgrade` | Call `upgrade` MCP tool, run returned shell command, display as checklist |
+| `ctx purge` | Call `purge` MCP tool with confirm: true. Warns before wiping knowledge base. |
+
+After /clear or /compact: knowledge base and session stats preserved. Use `ctx purge` to start fresh.
+
+## Windows notes
+
+**PowerShell cmdlets** — Sandbox uses bash. PowerShell cmdlets (`Format-List`, `Get-Culture`, etc.) fail with `command not found`. Wrap with `pwsh -NoProfile -Command "..."`.
+
+**Relative paths** — Sandbox CWD is temp dir, not project root. Convert to absolute paths. Ask user to confirm if unknown.
+
+**Windows drive letters** — Sandbox runs Git Bash / MSYS2. `X:\path` → `/x/path` (lowercase, no `/mnt/`). Never emit `/mnt/<letter>/`.
+
+**Quote paths** — Spaces in paths cause splits. Always double-quote: `rg "symbol" "$REPO_ROOT/some dir/Source"`.

package/configs/codex/config.toml

@@ -0,0 +1,8 @@
+# Codex CLI — context-mode MCP server configuration
+# Add this to ~/.codex/config.toml
+
+[features]
+hooks = true
+
+[mcp_servers.context-mode]
+command = "context-mode"

package/configs/codex/hooks.json

@@ -0,0 +1,47 @@
+{
+  "hooks": {
+    "PreToolUse": [
+      {
+        "matcher": "local_shell|shell|shell_command|exec_command|Bash|Shell|apply_patch|Edit|Write|grep_files|ctx_execute|ctx_execute_file|ctx_batch_execute|ctx_fetch_and_index|ctx_search|ctx_index|mcp__",
+        "hooks": [
+          { "type": "command", "command": "context-mode hook codex pretooluse" }
+        ]
+      }
+    ],
+    "PostToolUse": [
+      {
+        "hooks": [
+          { "type": "command", "command": "context-mode hook codex posttooluse" }
+        ]
+      }
+    ],
+    "SessionStart": [
+      {
+        "hooks": [
+          { "type": "command", "command": "context-mode hook codex sessionstart" }
+        ]
+      }
+    ],
+    "PreCompact": [
+      {
+        "hooks": [
+          { "type": "command", "command": "context-mode hook codex precompact" }
+        ]
+      }
+    ],
+    "UserPromptSubmit": [
+      {
+        "hooks": [
+          { "type": "command", "command": "context-mode hook codex userpromptsubmit" }
+        ]
+      }
+    ],
+    "Stop": [
+      {
+        "hooks": [
+          { "type": "command", "command": "context-mode hook codex stop" }
+        ]
+      }
+    ]
+  }
+}

package/configs/copilot-cli/.github/plugin/plugin.json

@@ -0,0 +1,23 @@
+{
+  "name": "context-mode",
+  "description": "context-mode for GitHub Copilot CLI: sandboxed code execution in 11 languages, an FTS5 knowledge base with BM25 ranking, and session capture. Saves your context window by keeping raw bytes out of the conversation.",
+  "version": "2.0.0",
+  "keywords": [
+    "mcp",
+    "context-window",
+    "sandbox",
+    "code-execution",
+    "fts5",
+    "bm25"
+  ],
+  "author": {
+    "name": "Mert Koseoğlu",
+    "url": "https://github.com/mxalbert1996"
+  },
+  "homepage": "https://github.com/mxalbert1996/context-mode#readme",
+  "repository": "https://github.com/mxalbert1996/context-mode",
+  "license": "Elastic-2.0",
+  "skills": [
+    "./skills/context-mode"
+  ]
+}

package/configs/copilot-cli/.mcp.json

@@ -0,0 +1,12 @@
+{
+  "mcpServers": {
+    "context-mode": {
+      "command": "context-mode",
+      "args": [],
+      "env": {
+        "CONTEXT_MODE_PLATFORM": "copilot-cli",
+        "CONTEXT_MODE_COPILOT_PLUGIN": "1"
+      }
+    }
+  }
+}