Skip to content

gemini-cli-extensions/sre-extension

v0.1.21Apache-2.0

πŸ‰ SRE extension for AI agents: incident investigation, SLO management, cloud monitoring and log auditing.

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.1.21] - 2026-07-31

Changed

  • Architecture Discovery Skill: Switched from external Mermaid rendering (mermaid.ink) to secure, offline Graphviz (dot) rendering for GCP architecture PNG generation to prevent data exfiltration.
  • Testing: Added test_render_architecture_png.py to test/run_tests.sh to validate local rendering script offline capabilities.

[0.1.20] - 2026-06-25

Added

  • GitHub Copilot CLI support: Added .copilot-plugin/plugin.json (Copilot-specific plugin manifest) and .copilot-plugin/marketplace.json (canonical Copilot marketplace location).
  • Justfile: New install-copilot (local dev install) and install-copilot-persistent (marketplace install) recipes.
  • Documentation: Updated INSTALL.md with a GitHub Copilot CLI installation section (Option A β€” local, Option B β€” marketplace). Updated README.md compatibility table and quickstart to include Copilot CLI.

[0.1.19] - 2026-06-15

Changed

  • GitHub Actions: Opted into Node.js 24 for all workflows by setting FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: 'true'. Updated actions/github-script to v8.0.0 to resolve deprecation warnings.

[0.1.18] - 2026-06-15

Fixed

  • Gemini CLI GHA Integration: Resolved issue #37 by correcting GitHub MCP tool names (get_issue, get_pull_request, etc.) in TOML prompts and workflow YAML configurations. Switched from Docker to npx for the GitHub MCP server to improve reliability and performance in GitHub Actions.

[0.1.17] - 2026-06-09

Changed

  • Documentation: Updated walkthrough video links and environment configurations for workspace trust.

[0.1.16] - 2026-06-05

Added

  • Architecture Discovery Skill: Added the gcp-architecture-discovery skill (skills/gcp-architecture-discovery/) to discover and map GCP infrastructure architecture including compute, networking, storage, and service dependencies.

Changed

  • Entrypoint & Setup Orchestration: Updated gcp-setup and investigation-entrypoint skills to integrate the new architecture discovery workflow as a mandatory baseline and incremental investigation step before performing queries or log analysis.

[0.1.15] - 2026-06-02

Changed

  • Outage Demo Video Thumbnail: Swapped the demo video preview to use a stunning custom screenshot (docs/sre-demo-video-thumbnail.png) showing the active multi-endpoint availability graph and SRE checklist for maximum visual impact.

[0.1.14] - 2026-06-02

Added

  • Outage Demo Video: Embedded a high-quality video link at the top of README.md showcasing a live SRE investigation and PostMortem generation with the extension.
  • Installation Walkthrough Video: Added a step-by-step video guide card at the top of INSTALL.md to help users get set up quickly.

[0.1.13] - 2026-06-01

Changed

  • Investigation Entrypoint: Added guidelines instructing the Investigator to automatically leverage sparklines and ASCII/Unicode graphs (export_timeseries_to_csv.py or csv_to_sparkline.py) alongside start/end timestamp context to give the user a rapid visual overview of incident metrics.

[0.1.12] - 2026-05-29

Added

  • Harness Compatibility Matrix: Introduced a comprehensive client capability and support matrix directly in README.md, clarifying feature coverage (installation, GCP MCP setup, and general SRE capabilities) across Gemini CLI, Antigravity CLI, Claude Code, and OpenAI Codex.

[0.1.11] - 2026-05-29

Changed

  • just install-agy: Enhanced the recipe to output a clean, green success emoji showing the installed version on first-time installation as well, matching the second-time check output.

[0.1.10] - 2026-05-28

Added

  • just plugin-version: Exposed a parameterized recipe that retrieves the version from any plugin JSON manifest file.
  • test/get_plugin_version.sh: Added a robust helper bash script to extract version info from any JSON file using jq (with a zero-dependency fallback to grep and sed).
  • OneMCP Support for Antigravity: Refactored gcp-mcp-setup skill and setup_onemcp.py to seamlessly configure modern Antigravity CLI and Editor paths.
  • Structured Reference Templates: Created structured settings.json (Gemini CLI) and mcp_config.json (Antigravity) schemas under skills/gcp-mcp-setup/references/.

Changed

  • just install-agy: Improved the recipe to check for an existing installation, compare the installed version with the workspace version, and print a warning or success message indicating whether they are in sync (e.g., 0.1.9 vs 0.1.10).
  • OneMCP Output Optimization: Modified the config generator to write tailored settings keys (httpUrl for Gemini and serverUrl for Antigravity) to prevent diagnostic key pollution.

[0.1.9] - 2026-05-28

Added

  • Automated Validation Tests: Created a unified test runner test/run_tests.sh that checks version consistency across manifests (test/check_manifest_versions.py) and ensures skills frontmatter conforms to specifications.
  • just test: Added a just test recipe to easily execute all validation checks with a single command.

Fixed

  • Skill Frontmatter: Quoted name/description fields, added missing emojis, and aligned statuses to satisfy metadata schema requirements across multiple skills.

[0.1.8] - 2026-05-28

Added

  • just install-agy: Added a just install-agy recipe to automate cloning the SRE extension repository directly into the agy plugins directory (~/.gemini/config/plugins/sre-extension).
  • just install-geminicli: Added a just install-geminicli recipe to install the extension via the Gemini CLI extensions command.
  • just install-claude: Added a just install-claude recipe to run Claude Code pointing to this plugin directory.

[0.1.7] - 2026-05-22

Added

  • Plugin Manifest: Added plugin.json at the root of the repository to support plugin integration and discovery as a named SRE extension.
  • Claude Plugin Support: Added .claude-plugin/plugin.json to enable native compatibility with Claude Code plugins.
  • Codex Plugin Support: Added .codex-plugin/plugin.json to enable native compatibility with OpenAI Codex plugins.
  • Documentation: Updated README.md and USER_MANUAL.md with installation and loading guides for Antigravity, Claude Code, and OpenAI Codex plugin systems.

Changed

  • Extension Config: Bumped extension metadata version to 0.1.7 in gemini-extension.json.

[0.1.6] - 2026-04-20

Added

  • Monitoring Graphs: Extracted csv_to_sparkline.py to autogenerate Unicode sparklines wrapped in pipes for better spacing visibility on arbitrary CSVs.

Changed

  • Cloud Monitoring: Refactored the terminal statistics to adopt the 'Sparkline' naming convention and use the pipe encapsulation framework.

[0.1.5] - 2026-04-17

Added

  • Cloud Monitoring: Added 'gist' / 'graphical shape' feature to visually represent time-series data.

[0.1.4] - 2026-04-17

Added

  • New Skill: postmortem-aggregator (v0.0.1) moved from gemini-cli-custom-commands.

Changed

  • Merged Skills: Consolidated postmortem-generator-copy and the external postmortem-generator into a single postmortem-generator skill (v0.0.5).
    • Improved the merged SKILL.md with combined IRM examples (Google IRM, ServiceNow, JIRA, etc.) and better execution instructions.
    • Standardized directory structure for postmortem output.

[0.1.3] - 2026-04-16

Added

  • GitHub Actions Integration:
    • Added workflows for issue triage, PR review, and automated planning/execution.
    • Configured custom Gemini CLI commands for GHA environment (gemini-invoke, gemini-plan-execute, etc.).

[0.1.2] - 2026-04-16

Changed

  • Moved policy.toml to policies/ directory for better organization.

[0.1.1] - 2026-04-16

Changed

  • Pre-publication Cleanup:
    • Replaced internal email addresses with GitHub profile links in README.md and skill metadata.
    • Removed internal go/ shortlinks and /google/ infrastructure paths.
    • Moved internal-only TODOs to untracked local files.
    • Restricted experimental agents/ directory from the public repository.
  • Improved Tooling:
    • Added a pre-publish-checker skill with automated scripts for scanning profanity, internal links, and professionalism.
    • Added a synopsis script for tracking publication readiness.

[0.1.0] - 2026-04-14

Added

  • New Skills:
    • cloud-logging: Analyze GCP JSON logs and convert to Apache format.
    • cloud-monitoring: Export time-series data to CSV and setup frontend SLOs.
    • monitoring-graphs: Generate high-quality annotated incident graphs using Python.
    • anomaly-detection: Detect anomalies using Isolation Forest, KNN, and Z-score.
    • data-ingestion: Merge and parse time-series data from various sources.
  • Validation Tools:
    • New test/check_skills_frontmatter.py script to enforce skill standards (names, descriptions, dragon emojis πŸ‰).
    • Added --no-warnings and --verbose flags to the skills validator.
    • Integrated justfile targets for linting and cross-project checks.
  • Documentation:
    • Added docs/sre-extension-logo.png and integrated it into the README.md.
    • Updated docs/SKILLS-STANDARDS.md with strict naming and frontmatter requirements.

Changed

  • Reorganized skills folder structure to use dashes instead of underscores.
  • Renamed postmortem-generator-copy to postmortem-generator after reconciliation.
  • Standardized all skill names and descriptions to pass strict frontmatter validation.

Removed

  • Cleaned up old migration documentation (docs/MIGRATION_META_PLAN.md, etc.).

[0.0.5] - 2026-04-13

Added

  • Added Apache 2.0 LICENSE file.
  • Moved docs/user-manual.md to root USER_MANUAL.md for better visibility.

[0.0.4] - 2026-04-02

Added

  • Initialized gcp-playbooks skill (v0.0.1) to provide SRE playbooks for GCP/GKE investigations.
  • Added CONTRIBUTING.md with open-source and Google best practices.
  • Documented cross-project discovery using gcloud asset search-all-resources.
  • Tracked by b/498511917 and b/498849864.

Changed

  • Refined GEMINI.md and README.md with updated contributor info and Gerrit best practices.
  • Reorganized gcp-playbooks reference structure for better determinism and DRYness.

[0.0.3] - 2026-03-30

Added

  • Added optional --google-maps-key flag to gcp-mcp-setup to configure google-maps MCP server.
  • Added identity consistency check (gcloud vs ADC) to gcp-mcp-setup verification script.

Changed

  • Refactored gcp-mcp-setup testing: consolidated bash scripts into a single comprehensive Python verification script (verify_setup.py).
  • Improved gcp-mcp-setup setup script to use explicit flags (--local / --global) for safer settings injection.

[0.0.2] - 2026-03-30

Added

  • Added OneMCP setup skill (v0.0.1) to automate the configuration of Google Managed MCP servers.
  • Migrated OneMCP setup script to Python for safer JSON injection.
  • Added BigQuery and Firestore MCPs back into the OneMCP setup script.

[0.0.1] - 2026-03-25

Added

  • Created the core Migration Meta Plan to move from internal tools to open-source GCP/GKE tooling.
  • Configured GCP/GKE specific guidelines and OneMCP constraints in GEMINI.md.
  • Initialized agents/outage-investigator.md tuned for Cloud Logging, Cloud Monitoring, and GKE cluster response.
  • Introduced a Mitigation Taxonomy for mapping incident mitigation strategies to GCP/GKE actions.
  • Included gemini-extension.json for custom script-based MCP integrations.
  • Integrated skaffold as the primary tool for testing codebase-level deployment mitigations.