gald3r

Triage non-code defects (spec_defect, policy_incongruity, design_gap): assess risk, auto-fix bounded safe issues, and log the rest as tracked bugs.

Compress the NON-gald3r sections of AGENTS.md/CLAUDE.md (and *memory*.md) to cut token overhead, while strictly preserving the install-managed gald3r SECTION ranges, code blocks, and URLs. Dry-run by default; apply only after confirmation.

Own and manage all feature data — FEATURES.md index, features/ individual files, staging lifecycle (staging→specced→committed→shipped), harvest source collection, and feature promotion. Single source of truth for everything feature-related.

Create well-structured git commits following gald3r conventions, with proper type prefixes, task references, and clean trailers (no AI co-author footers per C-021).

Tiered .gald3r/ health and intervention system. L1=triage (structural), L2=diagnosis (plan coherence), L3=surgery (cross-subsystem interface audit), L4=ecosystem (linked project negotiation). Replaces g-skl-medic.

gald3r_muninn MCP — query the local codebase knowledge graph for impact analysis, caller chains, dependencies, and symbol search. Clean-room rewrite (T1147 epic, T1153-T1158) of the GitNexus integration; auto-loaded by the example_app MCP server from docker/gald3r/tools/plugins/muninn/. Wire into g-go-code Step b0 Impact Scan before any implementation.

Policy-as-code guardrail — loads an org policy bundle (file offline, world_tree online) and exposes a CHECK op used by the pre-tool-call and pre-commit hooks to deterministically block/allow. Team/Org tier only; free/retail installs run an empty/default bundle.

Own and manage PROJECT.md (mission, goals, project linking). Single source of truth for project identity. Constraints operations delegated to g-skl-constraints.

Own and manage all release data — RELEASES.md index and releases/ individual files. Operations: CREATE new release, ASSIGN tasks to a release, STATUS summary, PUBLISH ROADMAP.md, ACCELERATE target dates with cascading shift to subsequent planned releases, SYNC reconcile CHANGELOG entries against release records (C-023).

Initialize gald3r in a project — folder structure and template files for task management.

Semantic versioning and release management for gald3r projects. Promotes CHANGELOG [Unreleased] to a versioned release, bumps VERSION, updates README badge, creates git tag, and optionally publishes a GitHub release. Also handles incremental CHANGELOG entries during development.

Show project status — session context, active tasks, phase progress, goals, ideas.

Own and manage SUBSYSTEMS.md (registry + mermaid graph) and subsystems/ spec files. Tracks what subsystems exist, their boundaries, dependencies, and activity logs.

Own and manage all task data — TASKS.md index, tasks/ individual files, status transitions, sync validation, complexity scoring, and sprint planning. Single source of truth for everything task-related.

Configurable product-tier onboarding skill. SETUP creates release_profiles/, scaffolds template_{tier}/ directories, and writes .gald3r/.identity tier metadata. ENABLE annotates existing SUBSYSTEMS.md with min_tier:, infers defaults from subsystem content, and calls platform_parity_sync -TierSync. Ships in full + adv tiers only (slim installs are pre-configured).

Workspace-Control Mode skill. Reads .gald3r/linking/workspace_manifest.yaml as the canonical registry and provides STATUS, VALIDATE, MEMBER LIST, SPAWN, ADOPT, EXPORT/SYNC dry-run planning, and member `.gald3r/` marker bootstrap/remediate/validate operations for manifest-declared repositories.

Register another project as a child of the current project. Creates or updates workspace/topology.md on both sides when the target is locally accessible.

File-only continual learning — agents self-report insights to project and vault memory notes. No JSONL scanning, no schedulers, no SentinelOne risk.

Spawn a new gald3r project from the current project. Creates the new project folder in the same ecosystem root, installs gald3r (matching the current project's install type — symlinks or fresh template), seeds it with any passed description/features/code, runs gald3r-setup, and immediately links both projects via WPAC topology (--parent | --sibling | --child).

Authoritative reference for Kiro CLI (Amazon's terminal agent, the Q Developer CLI rebrand) customization in gald3r projects. Covers .kiro/steering/ + AGENTS.md, Agent Skills (SKILL.md), JSON custom agents + subagents, lifecycle hooks, slash commands, MCP, and gald3r install verification. Distinct from Kiro IDE (g-skl-platform-kiro).

Authoritative reference for Pi (badlogic/pi-mono terminal coding-agent harness) customization in gald3r projects. Covers hierarchical AGENTS.md/CLAUDE.md instructions, native Agent Skills (SKILL.md), native prompt-template slash commands, TypeScript-extension lifecycle hooks (pi.on events), the absent MCP surface, and gald3r install verification.

Authoritative reference for ZCode (Z.ai / Zhipu, GLM-5.2 Agentic Development Environment) customization in gald3r projects. Covers the two-scope AGENTS.md convention (global-then-workspace, no imports/hierarchy), native Agent Skills (SKILL.md), native slash commands, Beta global-only subagents, native MCP, the absent hooks surface, and gald3r install verification.

<role in one line>

Triage non-code defects (spec_defect, policy_incongruity, design_gap): assess risk, auto-fix bounded safe issues, and log the rest as tracked bugs.

Production browser harness for agentic web tasks requiring persistent sessions, stealth/anti-detect, CAPTCHA solving, and self-healing CDP code generation via the browser-use library (YC W25, $17M seed). Use for login-required sites, anti-bot environments, competitive intel, and long-running multi-step web tasks.

SCA/license compliance scanning — wraps ORT, FOSSA, Snyk, and PMD CPD behind a unified gald3r interface with SCAN/REPORT/GATE/STATUS operations. Produces structured compliance reports with pass/warn/fail verdicts.

Engineering team delegation workflow skill — delegation patterns adapted for AI agent orchestration. Task briefs, code review request templates, quality gates, and handoff protocols.

Create well-structured git commits following gald3r conventions, with proper type prefixes, task references, and clean trailers (no AI co-author footers per C-021).

Render human-facing reports (status, review, backlog) as themed HTML using docs/templates/ + docs/themes/. Operations RENDER, CHOOSE_THEME, VALIDATE, EXPORT. Invoked by --html flag commands (T1318). Never used for coordination files (TASKS.md, BUGS.md, task specs) which are always markdown.

Emit gald3r command output (status, review, backlog) as structured JSON for scripting, CI gates, dashboards, and agent-to-agent handoff. Operations SERIALIZE, SCHEMA, VALIDATE, EXPORT. Invoked by --json flag commands (T1381). Mirrors g-skl-html-output. Coordination state files stay markdown.

User-invoked terse mode toggle. Suppresses the active personality pack, response structure scaffolding, and footer metadata so the agent returns bare signal only. Intended for debugging, rapid lookups, and high-volume Q&A sessions where the standard ceremony adds noise. Deactivates at session boundary or on explicit toggle-off.

Tiered .gald3r/ health and intervention system. L1=triage (structural), L2=diagnosis (plan coherence), L3=surgery (cross-subsystem interface audit), L4=ecosystem (linked project negotiation). Replaces g-skl-medic.

gald3r_muninn MCP — query the local codebase knowledge graph for impact analysis, caller chains, dependencies, and symbol search. Clean-room rewrite (T1147 epic, T1153-T1158) of the GitNexus integration; auto-loaded by the example_app MCP server from docker/gald3r/tools/plugins/muninn/. Wire into g-go-code Step b0 Impact Scan before any implementation.

Authoritative reference for Cursor IDE customization in gald3r projects. Covers .cursor/ folder layout, all supported primitives (rules/skills/agents/commands/hooks/MCP), AGENTS.md instruction file, .claude/.codex/.agents cross-tool reuse, parity tiers, and install verification.

Authoritative reference for Hermes Agent (Nous Research — self-improving CLI + Telegram/Discord/Slack gateway) customization in gald3r projects. Covers ~/.hermes/ config (config.yaml/.env), AGENTS.md-native instruction loading (auto-injected with SOUL.md/CLAUDE.md/.cursorrules), agentskills.io SKILL.md skills (the gald3r distribution opportunity), delegate_task subagents, MCP (mcp_servers), the native config.yaml hooks: surface (17+ events, pre_tool_call blocking), and gald3r install verification.

Authoritative reference for Kiro CLI (Amazon's terminal agent, the Q Developer CLI rebrand) customization in gald3r projects. Covers .kiro/steering/ + AGENTS.md, Agent Skills (SKILL.md), JSON custom agents + subagents, lifecycle hooks, slash commands, MCP, and gald3r install verification. Distinct from Kiro IDE (g-skl-platform-kiro).

Authoritative reference for the gald3r plugin system — install, remove, update, list, scaffold, and compatibility-check third-party plugins that bundle skills/commands/agents/rules/hooks. Documents the gald3r-plugin.yaml manifest, registry.json, the installed.yaml ledger, and the plugin_source: provenance convention (ADR-015 / SS-007). Single source of truth for everything plugin-related.

Documentation URL ingestion with periodic revisit. Crawls URLs into research/platforms/, tracks staleness per _index.yaml, surfaces stale count at session start. Depends on g-skl-crawl (crawl4ai).

YouTube transcript ingestion into the vault. Uses yt-dlp to fetch transcripts locally — no Docker, no MCP, no screen captures. Stores in research/videos/ with analysis_depth=transcript_only for future vision upgrade.

Initialize gald3r in a project — folder structure and template files for task management.