Propagates the 10-reviewer peer list across agent frontmatter, Position/Critical prose, shared-patterns, skill dispatchers, gate validators, and docs — resolving drift left behind when multi-tenant-reviewer and lib-commons-reviewer were added to the pool. Also fixes broken shared-pattern paths in lib-commons-reviewer and adds substantive blocker criteria to multi-tenant-reviewer plus codebase-context severity heuristic (Lerian third-rail vs external recommendation) to lib-commons-reviewer.
X-Lerian-Ref: 0x1
docs: align all documentation with the new 10-reviewer workflow
docs(lib-commons): update skill to reflect major v5.0.2 release
chore(hooks): tune context reminder hook to reduce token consumption
The cache-first feature injected <standards> blocks into 5 reviewer dispatch prompts, but only performance-reviewer parsed them — code, business-logic, security, and test reviewers ignored the injection and re-fetched standards on every run. The orchestrator-side half shipped without the consumer-side half.
Each of the 4 default reviewer agents now has a Standards Loading (Cache-First) section implementing the canonical three-tier resolution: cache hit from state.cached_standards, cache miss via WebFetch with warning, standalone fallback to hardcoded URLs. The per-reviewer fallback URL lists are unique and derived from each reviewer's standards_slice.
The 5-way verbatim duplication of the dispatch block in codereview/SKILL.md is collapsed to per-reviewer URL-slice markers that reference shared-patterns/standards-cache-protocol.md as the single source of truth. The shared pattern now carries the canonical <standards> template, orchestrator pseudocode, three-tier protocol, rolling-standards policy, and an anti-rationalization table defending cache-first behavior. Net 117-line reduction in codereview/SKILL.md.
X-Lerian-Ref: 0x1
Update all skills, agents, shared patterns, docs, and platform configs that still referenced 7 code review agents to reflect the 8th reviewer (ring:performance-reviewer). Includes agent self-descriptions, dispatch instructions, anti-rationalization tables, slicing math, gate tables, and opencode schema definitions.
X-Lerian-Ref: 0x1
feat(hooks): add schema validator and enhance skill generation
refactor(*): align all components with the new frontmatter schema
refactor(commands): replace arguments object with simple argument-hint
refactor(agents): remove invalid version field from agent frontmatter
test(hooks): add unit tests for frontmatter validation and generation
Introduces a new parallel reviewer, `ring:dead-code-reviewer`, to
systematically identify orphaned code resulting from changes. This
enhances codebase hygiene by detecting unused helpers, types, and
tests, thereby reducing technical debt and improving maintainability.
The parallel code review system is expanded from 6 to 7 concurrent
reviewers. All related workflows, skills, commands, and documentation
have been updated to integrate this new capability.
- Add the `ring:dead-code-reviewer` agent, which analyzes code across
three concentric rings (target, direct dependents, ripple effect) to
find unreachable code.
- Update the `/ring:codereview` command and `ring:requesting-code-review`
skill to dispatch all 7 reviewers in a single parallel step.
- Adjust project documentation, including architecture diagrams, manuals,
and agent guides, to reflect the 7-reviewer workflow.
Removes the hard-coded `model: "opus"` requirement from all agent
definitions, skill examples, and command documentation. The orchestrator
is now responsible for model selection, making the agent framework more
flexible and model-agnostic.
Key changes include:
- Removing the `model` parameter from all agent frontmatter and Task
tool examples.
- Deleting the "Model Requirement" hard gate and self-verification
sections from all agent prompts. This simplifies the prompts and
removes model-specific enforcement logic.
- Updating core documentation (ARCHITECTURE, MANUAL, README) to reflect
that model selection is an orchestrator concern.
- Removing redundant `changelog` and `last_updated` fields from agent
frontmatter, as this information is tracked by source control.
This architectural shift allows the orchestrator to dynamically choose
the best model for a given task based on complexity, cost, or speed,
without being constrained by the agent's definition. It simplifies
agent maintenance and prepares the system for easier integration of
future AI models.
Code reviews that focus only on changed files can miss how those
changes break other parts of the codebase. Callers, contract
consumers, and dependent modules can be inadvertently affected, leading to
bugs that are difficult to detect in a localized review.
This commit introduces a new agent, `ring:consequences-reviewer`, to
systematically trace these "ripple effects." Its purpose is to analyze
the downstream impact of code changes by walking caller chains and
verifying consumer contracts.
The core parallel code review workflow, orchestrated by the
`/ring:codereview` command, now dispatches six agents in parallel
instead of five. All related documentation, skills, and agent prompts
have been updated to reflect this more robust 6-reviewer process.
Move MUST/CANNOT/REQUIRED/MANDATORY to beginning of instruction sentences and lowercase context words (all/any/each/every) per lexical salience guidelines across business-logic, code, and test reviewers.
X-Lerian-Ref: 0x1
Expand business-logic-reviewer Standards Compliance Report with standards checklist, report template, and documentation rules. Wrap Pressure Resistance, When Not Needed, and Standards Compliance sections in XML semantic tags for code-reviewer and test-reviewer.
X-Lerian-Ref: 0x1
Wrap Pressure Resistance, When Not Needed, and Standards Compliance sections in XML semantic tags for business-logic and security reviewers. Replace weak minimal review phrasing with enforcement-first MANDATORY blocks across code, test reviewers. Add required tag to codebase-explorer standards handoff.
X-Lerian-Ref: 0x1
Add Pressure Resistance, When Not Needed, and Standards Compliance Report sections to business-logic-reviewer, code-reviewer, security-reviewer, and test-reviewer. Add Non-Negotiables to code-reviewer. Add Standards Compliance Report (N/A) to codebase-explorer. All sections follow nil-safety-reviewer reference pattern.
X-Lerian-Ref: 0x1
Transform all user-facing Ring component references from colon-based
notation (ring:) to a standardized hyphen-based format (ring-).
This change affects skill names, agent names, and slash commands in all
documentation, guides, and examples. The goal is to create a consistent
and user-friendly namespace that is easier to read and type.
For example:
- Skill/Agent references like `ring:code-reviewer` become `ring-code-reviewer`
- Slash commands like `/ring:codereview` become `/ring-codereview`
The colon-based format (`ring:`) is intentionally preserved in YAML
frontmatter (`name: ring:skill-name`) for internal metadata, separating
the user-facing display name from the internal tooling identifier.
Remove model: opus from frontmatter and Model Requirements sections
from all 7 default/ agents to allow model inheritance from Task invocation.
Affected agents:
- code-reviewer
- business-logic-reviewer
- security-reviewer
- test-reviewer
- nil-safety-reviewer
- write-plan
- codebase-explorer
X-Lerian-Ref: 0x1
This change introduces a namespacing convention for all built-in agents,
commands, and skills to prevent potential naming collisions with
user-defined entities.
All default entity names in the frontmatter are now prefixed with
'ring:'. For example, 'name: code-reviewer' becomes
'name: ring:code-reviewer'.
This creates a clear distinction between core framework components and
custom ones, improving the system's overall robustness and
extensibility.
This commit significantly expands the automated code review capabilities
by introducing two new specialized reviewers and refactoring the existing
agent architecture for improved maintainability and scalability.
The primary goal is to provide more comprehensive feedback by assessing
test quality and nil/null pointer safety, which are common sources of bugs.
To avoid massive duplication of instructions with the introduction of
new agents, common reviewer principles have been extracted into a new
`shared-patterns` directory. This makes each reviewer agent's prompt
more focused on its specific domain and ensures consistency across all
reviewers.
Changes:
- Add `test-reviewer` to analyze test quality, coverage, edge cases,
and anti-patterns.
- Add `nil-safety-reviewer` for Go and TypeScript to trace and report
nil/null pointer risks.
- Create a `shared-patterns` directory containing common logic for agent
model requirements, severity calibration, anti-rationalization,
pressure resistance, and output schemas.
- Refactor `code-reviewer`, `business-logic-reviewer`, and
`security-reviewer` to reference these shared patterns, significantly
reducing their file size and complexity.
- Update the `codereview` command and `requesting-code-review` skill to
orchestrate all five reviewers in parallel.
AI agents can generate plausible but flawed code ("AI slop") that
includes hallucinated dependencies, scope creep, and context-unaware
implementations. This is a primary source of bugs and security risks.
To address this, introduce a system-wide framework for detecting and
preventing AI slop, centered around a new shared skill document:
`ai-slop-detection.md`.
This framework is integrated across the agent ecosystem:
- Specialist agents (engineers, DevOps) now perform a mandatory
pre-submission self-check to catch their own slop before review.
- Reviewer agents (code, security, business) use new mandatory
checklists to enforce quality gates against slop from other agents.
- Validation agents (QA, SRE) must verify command outputs to prevent
reporting hallucinated metrics or test results.
This provides a defense-in-depth strategy against common AI failure
modes, improving code quality and mitigating supply-chain attack
vectors like "slopsquatting."
- Add shared-patterns/reviewer-orchestrator-boundary.md defining the
'Reviewers REPORT, Agents FIX' principle
- Add 'Orchestrator Boundary' section to all three reviewer agents:
- code-reviewer.md: CANNOT use Edit/Create tools, must report only
- business-logic-reviewer.md: CANNOT modify code, must report only
- security-reviewer.md: CANNOT fix vulnerabilities, must report only
- Include anti-rationalization tables for 'I'll just fix this' temptation
- Document workflow: reviewer reports -> orchestrator dispatches -> agent fixes
This ensures consistent separation of concerns across all reviewers.
Generated-by: Claude
AI-Model: claude-sonnet-4-20250514
Simplify naming convention by removing fully qualified prefixes from all components. Names now use simple format (e.g., code-reviewer instead of ring-default:code-reviewer).
X-Lerian-Ref: 0x1
Update all agent references to use the ring-{plugin}:{agent-name} format instead of bare agent names. This ensures consistent naming across all plugins and prevents ambiguity in multi-plugin environments.
X-Lerian-Ref: 0x1
Update all agents, skills, and commands across 9 plugins to use fully qualified names in YAML frontmatter. This enables unambiguous component identification in multi-plugin environments.
Changes:
- 42 agents: name field updated to ring-{plugin}:{agent-name}
- 87 skills: name field updated to ring-{plugin}:{skill-name}
- 27 commands: name field updated (added frontmatter where missing)
- CLAUDE.md: Fixed contradictory examples in naming conventions
- ARCHITECTURE.md: Updated invocation examples throughout
X-Lerian-Ref: 0x1
- Rename execution-report.md to output-execution-report.md
- Rename anti-rationalization.md to shared-anti-rationalization.md
- Rename orchestrator-principle.md to shared-orchestrator-principle.md
- Rename pressure-resistance.md to shared-pressure-resistance.md
- Rename red-flags.md to shared-red-flags.md
- Rename tdd-prompt-templates.md to template-tdd-prompts.md
- Update all references across skills, agents, and documentation
Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com>
docs(standards): mandate explicit model parameter for specialized agents
feat(reviewers): add guidance for skipping reviews on trivial changes
refactor(docs): simplify shared pattern files for better reusability
fix(security): add SSRF to blocker criteria and OWASP version check
fix(docs): align heading levels and table formats for consistency
This commit introduces a major overhaul to increase agent reliability,
predictability, and resistance to common failure modes. It enforces a
strict new standard for agent instructions across the entire system.
The "CLAUDE.md compliance" adds several mandatory sections to all agent
prompts, providing explicit, non-negotiable rules of engagement:
- `Blocker Criteria`: Defines what decisions an agent can and cannot make,
and when it must stop and escalate. Includes non-negotiable rules.
- `Severity Calibration`: Provides a strict rubric for classifying issue
severity to ensure consistency.
- `Pressure Resistance`: Equips agents with standard responses to resist
user pressure to skip steps or compromise quality.
- `Anti-Rationalization Table`: A critical defense mechanism that lists
common flawed heuristics and forces the agent to follow a required,
safer action instead.
refactor(skills): centralize shared patterns and use namespaced IDs
To support the agent updates and improve maintainability, this change
also refactors skill definitions.
New shared pattern files (`anti-rationalization.md` and
`pressure-resistance.md`) are created within each team's skills to
centralize common instructions and reduce prompt duplication.
Furthermore, all skill dependencies in `SKILL.md` files are updated to
use fully-qualified, namespaced IDs (e.g., `ring-default:writing-plans`).
This removes ambiguity in skill resolution and clarifies the dependency
graph between different teams and their respective skills.
Standardize agent invocation to use the format 'ring-{plugin}:{agent}'.
This change prevents potential name collisions between agents from
different plugins and makes agent ownership explicit. For example,
'code-reviewer' is now unambiguously 'ring-default:code-reviewer'.
All documentation, agent definitions, skills, and commands are updated
to reflect this new convention. The 'Task' tool parameter is also
updated from 'agent' to 'subagent_type' for consistency.
The built-in 'general-purpose' agent is exempt from this prefix.
Move the `pre-dev-feature` and `pre-dev-full` commands from the
`default` plugin to the `pm-team` plugin. This change aligns the
pre-development workflow with its logical product management function,
improving architectural clarity and modularity.
To support this multi-plugin structure, command invocation is updated
to a namespaced format: `/ring-{plugin}:{command}`. This prevents
potential command name collisions between different plugins.
fix(ralph-wiggum): Improve robustness of loop state and shell execution
Strengthen the shell script execution in `ralph-wiggum` commands by
encapsulating logic in `bash -c '...'`. This provides safer variable
expansion and prevents unintended globbing.
The session validation logic in the `stop-hook` is now more resilient,
handling varied quoting in the state file and failing securely if a
transcript path cannot be resolved.
feat(agent): introduce codebase-explorer for deep code analysis
feat(installer): dynamically load plugins in installer for easier updates
refactor(metadata): standardize skill and agent frontmatter for better tooling
The previous `sed`-based escaping was incomplete, only handling
backslashes and double quotes. This could generate invalid JSON if
instruction files contained control characters like tabs or carriage
returns, causing API requests to fail.
This commit replaces the `sed` command with a comprehensive `awk` script
that correctly escapes all necessary JSON control characters (`\"`, `\\`,
`\t`, `\r`, `\f`) and converts newlines to `\n` literals. This ensures
the generated JSON payload is RFC 8259 compliant and robust.
Additionally, the script is made executable, and changelogs are added
to the reviewer agent definitions for version tracking.
feat(plugin): introduce `ring-default` with core skills and agents
feat(hooks): add CLAUDE.md auto-generation and enhanced session start
docs: align documentation with new monorepo structure and vision
