auto-context-cli 1.0.3__tar.gz → 1.0.4__tar.gz

{auto_context_cli-1.0.3 → auto_context_cli-1.0.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: auto-context-cli
-Version: 1.0.3
+Version: 1.0.4
 Summary: Autonomous context & instruction manager for AI coding assistants
 Author: Arghya Biswas
 License: MIT
@@ -20,7 +20,7 @@ Description-Content-Type: text/markdown
 
 [![PyPI - Version](https://img.shields.io/pypi/v/auto-context-cli.svg?color=blue)](https://pypi.org/project/auto-context-cli/)
 [![PyPI - Python Version](https://img.shields.io/pypi/pyversions/auto-context-cli.svg)](https://pypi.org/project/auto-context-cli/)
-[![GitHub Release](https://img.shields.io/github/v/release/Cyber-arghya/auto-context?color=green)](https://github.com/Cyber-arghya/auto-context/releases/tag/v1.0.3)
+[![GitHub Release](https://img.shields.io/github/v/release/Cyber-arghya/auto-context?color=green)](https://github.com/Cyber-arghya/auto-context/releases/tag/v1.0.4)
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
 [![Platform](https://img.shields.io/badge/Platform-Windows%20%7C%20macOS%20%7C%20Linux%20WSL2-success.svg)](#-quick-one-line-install)
 [![AI Supported](https://img.shields.io/badge/AI%20Supported-Antigravity%20%7C%20Cursor%20%7C%20Claude%20Code%20%7C%20Copilot%20%7C%20Windsurf-orange.svg)](#-supported-ai-assistants)

{auto_context_cli-1.0.3 → auto_context_cli-1.0.4}/README.md

@@ -6,7 +6,7 @@
 
 [![PyPI - Version](https://img.shields.io/pypi/v/auto-context-cli.svg?color=blue)](https://pypi.org/project/auto-context-cli/)
 [![PyPI - Python Version](https://img.shields.io/pypi/pyversions/auto-context-cli.svg)](https://pypi.org/project/auto-context-cli/)
-[![GitHub Release](https://img.shields.io/github/v/release/Cyber-arghya/auto-context?color=green)](https://github.com/Cyber-arghya/auto-context/releases/tag/v1.0.3)
+[![GitHub Release](https://img.shields.io/github/v/release/Cyber-arghya/auto-context?color=green)](https://github.com/Cyber-arghya/auto-context/releases/tag/v1.0.4)
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
 [![Platform](https://img.shields.io/badge/Platform-Windows%20%7C%20macOS%20%7C%20Linux%20WSL2-success.svg)](#-quick-one-line-install)
 [![AI Supported](https://img.shields.io/badge/AI%20Supported-Antigravity%20%7C%20Cursor%20%7C%20Claude%20Code%20%7C%20Copilot%20%7C%20Windsurf-orange.svg)](#-supported-ai-assistants)

{auto_context_cli-1.0.3 → auto_context_cli-1.0.4}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "auto-context-cli"
-version = "1.0.3"
+version = "1.0.4"
 description = "Autonomous context & instruction manager for AI coding assistants"
 readme = "README.md"
 requires-python = ">=3.11"

{auto_context_cli-1.0.3 → auto_context_cli-1.0.4}/src/auto_context/__init__.py

@@ -1,3 +1,3 @@
 """auto-context: Autonomous context & instruction manager for AI coding assistants."""
 
-__version__ = "1.0.3"
+__version__ = "1.0.4"

{auto_context_cli-1.0.3 → auto_context_cli-1.0.4}/src/auto_context/commands/scan_cmd.py

@@ -131,3 +131,4 @@ def scan(
         f"• Hub size (context.md): [bold green]< 60 lines (Token Optimized)[/]",
         border_style="green",
     ))
+

{auto_context_cli-1.0.3 → auto_context_cli-1.0.4}/src/auto_context/main.py

@@ -69,7 +69,7 @@ def root(
             try:
                 v = pkg_version("auto-context")
             except Exception:
-                v = "1.0.3"
+                v = "1.0.4"
         console.print(f"init-context v{v}")
         raise typer.Exit()
 

auto_context_cli-1.0.4/src/auto_context/templates/AGENTS.template.md

@@ -0,0 +1,47 @@
+# Standing Agent Rules for {{PROJECT_NAME}}
+
+## Critical Mandate: Autonomous Context Synchronization & Precision Architecture
+As an AI coding assistant working in this repository (`{{PROJECT_DIR}}`), you are **strictly mandated** to maintain the Hub & Satellite context architecture, navigate codebases with laser precision, and autonomously crystallize new rules and skills. The user must **never** need to remind or instruct you.
+
+### 1. Precision Context Navigation Protocol
+Whenever exploring code or resolving tasks during an active session, follow the precision formula:
+$$\text{Precision Context} = \text{Route Map (Macro)} + \text{Grep/Symbol (Micro)} + \text{Targeted Range Read}$$
+- **Never bulk-read entire directories or loop through 50+ source files.**
+- First inspect `.agents/context/routes.md` or Section 2 of `context.md` for high-level file locations.
+- Use `grep` or symbol search to pinpoint exact files and line numbers.
+- Read only the specific target line range (e.g. `#L30-L70`) rather than loading entire large files.
+
+### 2. Mandatory Post-Action Protocol & Satellite Sync (Phase 3)
+Before concluding any turn or session, the agent MUST inspect and update context:
+- **Routes & Pages:** If any API endpoint or UI view is created, modified, or removed, immediately update `.agents/context/routes.md` (or run `init-context scan`).
+- **Database Models:** If any schema, migration, or entity is changed, immediately update `.agents/context/schema.md`.
+- **Architecture Gotchas:** If a non-obvious trap, port collision, or bug invariant is discovered, immediately record it in `.agents/context/gotchas.md`.
+- **Core Hub:** Keep `context.md` compact (< 400 tokens / 60 lines) as an executive routing hub.
+
+### 3. Domain Consolidation & Anti-Fragmentation (Single Responsibility)
+When codifying new guidelines or workflows, DO NOT create fragmented micro-files if a domain file already exists:
+- **Docker & Containerization:** Update/append to `.agents/skills/docker-ops/SKILL.md`.
+- **Git & Release Workflows:** Update/append to `.agents/skills/git-workflow/SKILL.md`.
+- **Coding Standards & Patterns:** Update/append to `.agents/rules/code-standards.md`.
+- **Security Boundaries:** Update/append to `.agents/rules/security-boundaries.md`.
+- Only create a new rule or skill file (`init-context -r` / `-s`) if it represents a distinctly new, independent domain.
+
+### 4. Context Distillation over Blind Deletion
+Never blindly delete or truncate context:
+- When problem logs or debugging notes grow beyond 50 lines, distill and synthesize the technical root causes into high-density invariants inside `.agents/context/gotchas.md`.
+- Preserve 100% of the architectural wisdom with 90% fewer tokens.
+
+### 5. Local Sandbox Test Credentials (`context.local.md`)
+- **Chat Continuity:** When starting a new session or switching chats, always read `context.local.md` to identify local dev ports, seed users, and mock passwords (`admin@test.local`) before running browser automation or curl tests.
+- **Strict Boundary:** Real production secrets, payment keys, and cloud tokens belong strictly in `.env`.
+- Ensure `context.local.md` is protected with `.gitignore` and `chmod 600`.
+
+### 6. Autonomous Learning & Tool Access
+- **Zero Token Waste:** DO NOT burn tokens manually searching directories to check for duplicates. Directly run `init-context -r <name> [-g]` — the CLI has instant built-in duplicate detection and collision guards.
+- **Mandatory Global Scope:** When the user mentions "for future", "always", or defines a reusable standard, **MANDATORY: Always pass `--global` (`-g`)** and immediately follow with `init-context memory push`.
+- Available CLI commands:
+  - `init-context scan`: Refresh stack-adaptive satellites (`routes.md`, `schema.md`, `gotchas.md`).
+  - `init-context -r <name> [-g]`: Create rule.
+  - `init-context -s <name> [-g]`: Create skill.
+  - `init-context update`: Safely sync public rules and upstream capabilities (non-breaking).
+  - `init-context memory push`: Push updates to central memory.

auto_context_cli-1.0.4/src/auto_context/templates/GEMINI.template.md

@@ -0,0 +1,47 @@
+# Standing Agent Rules for {{PROJECT_NAME}}
+
+## Critical Mandate: Autonomous Context Synchronization & Precision Architecture
+As an AI coding assistant working in this repository (`{{PROJECT_DIR}}`), you are **strictly mandated** to maintain the Hub & Satellite context architecture, navigate codebases with laser precision, and autonomously crystallize new rules and skills. The user must **never** need to remind or instruct you.
+
+### 1. Precision Context Navigation Protocol
+Whenever exploring code or resolving tasks during an active session, follow the precision formula:
+$$\text{Precision Context} = \text{Route Map (Macro)} + \text{Grep/Symbol (Micro)} + \text{Targeted Range Read}$$
+- **Never bulk-read entire directories or loop through 50+ source files.**
+- First inspect `.agents/context/routes.md` or Section 2 of `context.md` for high-level file locations.
+- Use `grep` or symbol search to pinpoint exact files and line numbers.
+- Read only the specific target line range (e.g. `#L30-L70`) rather than loading entire large files.
+
+### 2. Mandatory Post-Action Protocol & Satellite Sync (Phase 3)
+Before concluding any turn or session, the agent MUST inspect and update context:
+- **Routes & Pages:** If any API endpoint or UI view is created, modified, or removed, immediately update `.agents/context/routes.md` (or run `init-context scan`).
+- **Database Models:** If any schema, migration, or entity is changed, immediately update `.agents/context/schema.md`.
+- **Architecture Gotchas:** If a non-obvious trap, port collision, or bug invariant is discovered, immediately record it in `.agents/context/gotchas.md`.
+- **Core Hub:** Keep `context.md` compact (< 400 tokens / 60 lines) as an executive routing hub.
+
+### 3. Domain Consolidation & Anti-Fragmentation (Single Responsibility)
+When codifying new guidelines or workflows, DO NOT create fragmented micro-files if a domain file already exists:
+- **Docker & Containerization:** Update/append to `.agents/skills/docker-ops/SKILL.md`.
+- **Git & Release Workflows:** Update/append to `.agents/skills/git-workflow/SKILL.md`.
+- **Coding Standards & Patterns:** Update/append to `.agents/rules/code-standards.md`.
+- **Security Boundaries:** Update/append to `.agents/rules/security-boundaries.md`.
+- Only create a new rule or skill file (`init-context -r` / `-s`) if it represents a distinctly new, independent domain.
+
+### 4. Context Distillation over Blind Deletion
+Never blindly delete or truncate context:
+- When problem logs or debugging notes grow beyond 50 lines, distill and synthesize the technical root causes into high-density invariants inside `.agents/context/gotchas.md`.
+- Preserve 100% of the architectural wisdom with 90% fewer tokens.
+
+### 5. Local Sandbox Test Credentials (`context.local.md`)
+- **Chat Continuity:** When starting a new session or switching chats, always read `context.local.md` to identify local dev ports, seed users, and mock passwords (`admin@test.local`) before running browser automation or curl tests.
+- **Strict Boundary:** Real production secrets, payment keys, and cloud tokens belong strictly in `.env`.
+- Ensure `context.local.md` is protected with `.gitignore` and `chmod 600`.
+
+### 6. Autonomous Learning & Tool Access
+- **Zero Token Waste:** DO NOT burn tokens manually searching directories to check for duplicates. Directly run `init-context -r <name> [-g]` — the CLI has instant built-in duplicate detection and collision guards.
+- **Mandatory Global Scope:** When the user mentions "for future", "always", or defines a reusable standard, **MANDATORY: Always pass `--global` (`-g`)** and immediately follow with `init-context memory push`.
+- Available CLI commands:
+  - `init-context scan`: Refresh stack-adaptive satellites (`routes.md`, `schema.md`, `gotchas.md`).
+  - `init-context -r <name> [-g]`: Create rule.
+  - `init-context -s <name> [-g]`: Create skill.
+  - `init-context update`: Safely sync public rules and upstream capabilities (non-breaking).
+  - `init-context memory push`: Push updates to central memory.

auto_context_cli-1.0.4/src/auto_context/templates/maintain_context.template.md

@@ -0,0 +1,48 @@
+---
+name: maintain_context
+description: Mandates autonomous, continuous synchronization of context.md and proactive rule/skill learning on any repository or server configuration change without requiring user prompting.
+trigger: always_on
+---
+
+# Autonomous Context Maintenance & Self-Evolution Rule
+
+## 1. Precision Context Navigation Protocol
+Whenever exploring code or resolving tasks during an active session, follow the precision formula:
+$$\text{Precision Context} = \text{Route Map (Macro)} + \text{Grep/Symbol (Micro)} + \text{Targeted Range Read}$$
+- **Never bulk-read entire directories or loop through 50+ source files.**
+- First inspect `.agents/context/routes.md` or Section 2 of `context.md` for high-level file locations.
+- Use `grep` or symbol search to pinpoint exact files and line numbers.
+- Read only the specific target line range (e.g. `#L30-L70`) rather than loading entire large files.
+
+## 2. Mandatory Post-Action Protocol & Satellite Sync (Phase 3)
+Before concluding any turn or session, the agent MUST inspect and update context:
+- **Routes & Pages:** If any API endpoint or UI view is created, modified, or removed, immediately update `.agents/context/routes.md` (or run `init-context scan`).
+- **Database Models:** If any schema, migration, or entity is changed, immediately update `.agents/context/schema.md`.
+- **Architecture Gotchas:** If a non-obvious trap, port collision, or bug invariant is discovered, immediately record it in `.agents/context/gotchas.md`.
+- **Core Hub:** Keep `context.md` compact (< 400 tokens / 60 lines) as an executive routing hub.
+
+## 3. Domain Consolidation & Anti-Fragmentation (Single Responsibility)
+When codifying new guidelines or workflows, DO NOT create fragmented micro-files if a domain file already exists:
+- **Docker & Containerization:** Update/append to `.agents/skills/docker-ops/SKILL.md`.
+- **Git & Release Workflows:** Update/append to `.agents/skills/git-workflow/SKILL.md`.
+- **Coding Standards & Patterns:** Update/append to `.agents/rules/code-standards.md`.
+- **Security Boundaries:** Update/append to `.agents/rules/security-boundaries.md`.
+- Only create a new rule or skill file (`init-context -r` / `-s`) if it represents a distinctly new, independent domain.
+
+## 4. Context Distillation over Blind Deletion
+Never blindly delete or truncate context:
+- When problem logs or debugging notes grow beyond 50 lines, distill and synthesize the technical root causes into high-density invariants inside `.agents/context/gotchas.md`.
+- Preserve 100% of the architectural wisdom with 90% fewer tokens.
+
+## 5. Local Sandbox Test Credentials (`context.local.md`)
+- **Chat Continuity:** When starting a new session or switching chats, always read `context.local.md` to identify local dev ports, seed users, and mock passwords (`admin@test.local`) before running browser automation or curl tests.
+- **Strict Boundary:** Real production secrets, payment keys, and cloud tokens belong strictly in `.env`.
+- Ensure `context.local.md` is protected with `.gitignore` and `chmod 600`.
+
+## 6. Autonomous Learning & Tooling
+- **Zero Token Waste:** DO NOT burn tokens manually searching directories to check for duplicates. Directly run `init-context -r <name> [-g]` — the CLI has instant built-in duplicate detection and collision guards.
+- **Mandatory Global Scope:** When the user mentions "for future", "always", or defines a reusable standard, **MANDATORY: Always pass `--global` (`-g`)** and immediately follow with `init-context memory push`.
+- The agent has full access to run `init-context`:
+  - `init-context scan`: Refresh stack-adaptive satellites (`routes.md`, `schema.md`, `gotchas.md`).
+  - `init-context update`: Safely sync public rules and upstream capabilities (non-breaking).
+  - `init-context memory push`: Push updates to central memory.

{auto_context_cli-1.0.3 → auto_context_cli-1.0.4}/src/auto_context/templates/rules/hub-satellite-architecture.md

@@ -31,3 +31,4 @@ Prevent monolithic context file bloat, eliminate token waste, ensure 100% idempo
 4. **KISS & DRY Heuristics:**
    - Prefer language-agnostic filesystem convention heuristics (directories, manifests) over heavy language AST parsers that crash on in-flight syntax errors.
    - All scanners must degrade gracefully with zero unhandled exceptions.
+

auto_context_cli-1.0.3/src/auto_context/templates/AGENTS.template.md

@@ -1,53 +0,0 @@
-# Standing Agent Rules for {{PROJECT_NAME}}
-
-## Critical Mandate: Autonomous `context.md` Synchronization & Self-Evolution
-As an AI coding assistant working in this repository (`{{PROJECT_DIR}}`), you are **strictly mandated** to keep `{{PROJECT_DIR}}/context.md` continuously up to date and autonomously learn, persist, and crystallize new rules and skills. The user must **never** need to remind or instruct you.
-
-### 1. Mandatory Post-Action Protocol
-Whenever ANY of the following events occur in any session or conversation turn:
-- A configuration file or script is created, edited, moved, or deleted.
-- A problem, bug, or error is encountered and diagnosed.
-- Credentials, tokens, user accounts, or security keys are generated or updated.
-- Architectural decisions or operational steps are modified.
-
-**You MUST immediately update `{{PROJECT_DIR}}/context.md` before finalizing your response.**
-
-### 2. Autonomous Learning & Self-Evolution Protocol
-Whenever you learn a new pattern, receive a correction, observe a coding preference, or resolve a complex repeatable problem:
-1. **Identify the Type:**
-   - **Rule:** If it is a constraint, coding standard, safety boundary, or architectural preference -> Create a Rule (`init-context -r <name>`).
-   - **Skill:** If it is a multi-step procedure, deployment workflow, diagnosis runbook, or CLI sequence -> Create a Skill (`init-context -s <name>`).
-2. **Zero Token Waste:**
-   - **DO NOT burn tokens manually searching directories or reading files to check for existing rules.** Directly run `init-context -r <name> [-g]` — the CLI has instant built-in duplicate detection and collision guards.
-3. **Mandatory Global Scope:**
-   - When the user mentions "for future", "always", or defines a reusable standard, **MANDATORY: Always pass `--global` (`-g`)** and immediately follow with `init-context memory push`.
-   - **Project-Specific:** Only omit `-g` if strictly confined to this specific project repository.
-4. **Execution:**
-   - Use the `init-context` CLI tool to scaffold the file.
-   - Populate it with clear objectives, constraints, or step-by-step instructions.
-   - Inform the user of the new rule or skill codified.
-
-### 3. Tool Access: `init-context`
-You have native access to run `init-context` via terminal commands:
-- `init-context`: Scaffolds or repairs project context.
-- `init-context -r <rule-name> [-g]`: Creates a new rule with frontmatter.
-- `init-context -s <skill-name> [-g]`: Creates a new skill with frontmatter.
-- `init-context update`: Safely syncs public rules and upstream capabilities (non-breaking).
-- `init-context memory push`: Synchronizes central brain memory to private Git repository.
-
-### 4. Required Content & Structure in `context.md`
-1. **Master Credentials & Configuration Vault (Section 1):** Keep all ports, IPs, secrets, tokens, and endpoints synchronized.
-2. **Chronological Problem Log & Technical Solutions (Section 2):** Document symptom, root cause, exact commands/code used, and update Mermaid diagram.
-3. **Architecture & Component Layout (Section 3):** Keep stack overview and directory maps accurate.
-4. **Operations Runbook (Section 4):** Ensure 1-step reproduction instructions remain functional.
-
-### 5. Local Sandbox Credentials & Dummy Test Fixtures (`context.local.md`)
-- **Inspection Mandate:** When testing local dev servers, performing browser automation (Playwright/Puppeteer), or authenticating API endpoints, ALWAYS read `context.local.md` first before prompting the user for credentials.
-- **Strict Mock Boundary:** ONLY store dummy test credentials, mock accounts (`admin@test.local`), seed users, and local dev ports in `context.local.md`.
-- **Zero Real Secrets:** Real production secrets, payment keys, or personal tokens MUST NEVER be stored in `context.local.md` or `context.md`; reference them via `.env` or system environment variables.
-- **Git Protection:** Ensure `context.local.md` remains strictly ignored by Git (`.gitignore`) and protected with `chmod 600`.
-
-### 6. Two-Tier Progressive Disclosure Protocol (Token Optimization)
-- **Initial Boot Budget:** NEVER read full `SKILL.md` files into the initial context. Use Section 5 of `context.md` for lightweight routing (< 800 tokens).
-- **Dynamic On-Demand Loading:** Only when a user prompt matches a skill's triggers (e.g. "docker", "rebase", "diagnostics", "browser"), dynamically inspect the file using `read_file`.
-- **Zero Session Bloat:** Once the task concludes, do not keep the full skill body in persistent memory.

auto_context_cli-1.0.3/src/auto_context/templates/GEMINI.template.md

@@ -1,53 +0,0 @@
-# Standing Agent Rules for {{PROJECT_NAME}}
-
-## Critical Mandate: Autonomous `context.md` Synchronization & Self-Evolution
-As an AI coding assistant working in this repository (`{{PROJECT_DIR}}`), you are **strictly mandated** to keep `{{PROJECT_DIR}}/context.md` continuously up to date and autonomously learn, persist, and crystallize new rules and skills. The user must **never** need to remind or instruct you.
-
-### 1. Mandatory Post-Action Protocol
-Whenever ANY of the following events occur in any session or conversation turn:
-- A configuration file or script is created, edited, moved, or deleted.
-- A problem, bug, or error is encountered and diagnosed.
-- Credentials, tokens, user accounts, or security keys are generated or updated.
-- Architectural decisions or operational steps are modified.
-
-**You MUST immediately update `{{PROJECT_DIR}}/context.md` before finalizing your response.**
-
-### 2. Autonomous Learning & Self-Evolution Protocol
-Whenever you learn a new pattern, receive a correction, observe a coding preference, or resolve a complex repeatable problem:
-1. **Identify the Type:**
-   - **Rule:** If it is a constraint, coding standard, safety boundary, or architectural preference -> Create a Rule (`init-context -r <name>`).
-   - **Skill:** If it is a multi-step procedure, deployment workflow, diagnosis runbook, or CLI sequence -> Create a Skill (`init-context -s <name>`).
-2. **Zero Token Waste:**
-   - **DO NOT burn tokens manually searching directories or reading files to check for existing rules.** Directly run `init-context -r <name> [-g]` — the CLI has instant built-in duplicate detection and collision guards.
-3. **Mandatory Global Scope:**
-   - When the user mentions "for future", "always", or defines a reusable standard, **MANDATORY: Always pass `--global` (`-g`)** and immediately follow with `init-context memory push`.
-   - **Project-Specific:** Only omit `-g` if strictly confined to this specific project repository.
-4. **Execution:**
-   - Use the `init-context` CLI tool to scaffold the file.
-   - Populate it with clear objectives, constraints, or step-by-step instructions.
-   - Inform the user of the new rule or skill codified.
-
-### 3. Tool Access: `init-context`
-You have native access to run `init-context` via terminal commands:
-- `init-context`: Scaffolds or repairs project context.
-- `init-context -r <rule-name> [-g]`: Creates a new rule with frontmatter.
-- `init-context -s <skill-name> [-g]`: Creates a new skill with frontmatter.
-- `init-context update`: Safely syncs public rules and upstream capabilities (non-breaking).
-- `init-context memory push`: Synchronizes central brain memory to private Git repository.
-
-### 4. Required Content & Structure in `context.md`
-1. **Master Credentials & Configuration Vault (Section 1):** Keep all ports, IPs, secrets, tokens, and endpoints synchronized.
-2. **Chronological Problem Log & Technical Solutions (Section 2):** Document symptom, root cause, exact commands/code used, and update Mermaid diagram.
-3. **Architecture & Component Layout (Section 3):** Keep stack overview and directory maps accurate.
-4. **Operations Runbook (Section 4):** Ensure 1-step reproduction instructions remain functional.
-
-### 5. Local Sandbox Credentials & Dummy Test Fixtures (`context.local.md`)
-- **Inspection Mandate:** When testing local dev servers, performing browser automation (Playwright/Puppeteer), or authenticating API endpoints, ALWAYS read `context.local.md` first before prompting the user for credentials.
-- **Strict Mock Boundary:** ONLY store dummy test credentials, mock accounts (`admin@test.local`), seed users, and local dev ports in `context.local.md`.
-- **Zero Real Secrets:** Real production secrets, payment keys, or personal tokens MUST NEVER be stored in `context.local.md` or `context.md`; reference them via `.env` or system environment variables.
-- **Git Protection:** Ensure `context.local.md` remains strictly ignored by Git (`.gitignore`) and protected with `chmod 600`.
-
-### 6. Two-Tier Progressive Disclosure Protocol (Token Optimization)
-- **Initial Boot Budget:** NEVER read full `SKILL.md` files into the initial context. Use Section 5 of `context.md` for lightweight routing (< 800 tokens).
-- **Dynamic On-Demand Loading:** Only when a user prompt matches a skill's triggers (e.g. "docker", "rebase", "diagnostics", "browser"), dynamically inspect the file using `read_file`.
-- **Zero Session Bloat:** Once the task concludes, do not keep the full skill body in persistent memory.

auto_context_cli-1.0.3/src/auto_context/templates/maintain_context.template.md

@@ -1,37 +0,0 @@
----
-name: maintain_context
-description: Mandates autonomous, continuous synchronization of context.md and proactive rule/skill learning on any repository or server configuration change without requiring user prompting.
-trigger: always_on
----
-
-# Autonomous Context Maintenance & Self-Evolution Rule
-
-1. **Mandatory Synchronization**: Every agent working in this workspace MUST inspect and update `context.md` before finishing their turn whenever any of the following occur:
-   - Script or configuration edits (environment variables, docker-compose, configs, network parameters).
-   - Problem diagnosis, debugging steps, and technical solutions.
-   - User credentials, tokens, secrets, or permissions changes.
-   - Architectural decisions or operational runbooks.
-
-2. **Autonomous Learning & Skill Crystallization**:
-   - **Zero Token Waste:** DO NOT burn tokens manually searching directories or reading files to check for existing rules. Directly run `init-context -r <name> [-g]` — the CLI has instant built-in duplicate detection and collision guards.
-   - **Mandatory Global Scope:** When the user mentions "for future", "always", or defines a reusable standard, **MANDATORY: Always pass `--global` (`-g`)** and immediately follow with `init-context memory push`.
-   - Populates the file with clear instructions and notifies the user.
-
-3. **CLI Tool Availability**:
-   - The agent has full access to run `init-context` directly via terminal:
-     - `init-context`: Initialize/repair context.
-     - `init-context -r <name> [-g]`: Create rule.
-     - `init-context -s <name> [-g]`: Create skill.
-     - `init-context update`: Safely sync public rules and upstream capabilities (non-breaking).
-     - `init-context memory push`: Push updates to central memory.
-
-4. **No User Reminder Needed**: The user must never need to ask the agent to update `context.md` or learn rules. This is a strict automated post-condition for all actions.
-
-5. **Local Sandbox Test Credentials (`context.local.md`)**:
-   - For local dev logins, automated browser testing, and mock fixtures, inspect and update `context.local.md`.
-   - Never store real production secrets or payment tokens in markdown files. Real secrets belong in `.env`.
-   - Ensure `context.local.md` is strictly git-ignored and protected with `chmod 600`.
-
-6. **Two-Tier Progressive Disclosure**:
-   - Never load full `SKILL.md` bodies into the initial context. Use Section 5 of `context.md` for routing.
-   - Dynamically load specific skills via `read_file` only when a user prompt matches their triggers.