SkillsHooksPromptsAgentsPersonasModelsPoliciesToolsTemplatesBundlesCategoriesStart here
← All categories

category · governance

Governance & Meta 99

Sk 18 skillsHk 11 hooksPr 32 promptsAg 1 agentPe 5 personasPo 28 policiesTe 4 templatesJSON ↗
stable

Amendment Author

Guided workflow for authoring and publishing governance amendments against the four canonical files (Constitution, Common, Code, Writing). Walks the full lifecycle: identify the violation or gap, draft the amendment stub, review it against the governance model, apply it to the governance file, and optionally publish it as a versioned amendment atom. Triggers on: /amendment-author, 'author an amendment', 'create an amendment', 'propose a governance change'.

invoke
/amendment-author
skill/amendment-authorv1.0.0ai-atoms
stable

Atom Fleet Status

Full fleet snapshot of all *-atoms.com catalog sites — deploy health, DNS, HTTPS reachability, atom_types per catalog with item counts, schema compliance, /ai/index.json, Terraform backend, sidenav, brand asset uniformity, builder script, federation field, and registry presence. Use when the user asks about the state of the atoms fleet, wants to verify deployments, asks for a fleet snapshot or schema compliance report, or before starting any cross-fleet work.

invoke
/atom-status
skill/atom-statev1.0.0ai-atoms
stable

Atom Publisher

Guided 5-step workflow for drafting, canonicalizing, verifying, and publishing a new atom to the appropriate *-atoms catalog. Supports skill, hook, persona, profile, plugin, amendment, and action atom types. Routes to the correct catalog based on type. Triggers on: /atom-publisher, 'publish an atom', 'create an atom', 'contribute an atom'.

invoke
/atom-publisher
skill/atom-publisherv1.0.0ai-atoms
stable

Atoms Discover

Navigate the convergent-systems atom ecosystem to find, fetch, or scaffold atoms. Routes 'apply brand X' to brand-atoms, 'use persona Y' to persona-atoms, 'load skill Z' to skill-atoms, etc. Covers discovery, cross-atom reference resolution, and new-atom scaffolding.

invoke
/atoms-discover
skill/atoms-discoverv1.0.0ai-atoms
stable

Dispatch Parallel Agents

Decompose a multi-task problem into independent units and dispatch subagents in parallel. Use when facing 2+ independent tasks with no shared state or sequential dependencies. Each agent gets a self-contained prompt with full context.

skill/dispatching-parallel-agentsv1.0.0ai-atoms
stable

Hook Author

Guided workflow for authoring and installing governance hooks to the ~/.ai/hooks/ enforcement plane. Covers the full lifecycle: describe the behavior → write the script → validate → install → test. Supported events: PreToolUse, PostToolUse, Stop, SubagentStop, PreCompact, UserPromptSubmit. Triggers on: /hook-author, 'author a hook', 'create a hook', 'write a governance hook'.

invoke
/hook-author
skill/hook-authorv1.0.0ai-atoms
stable

Make Atoms

Discover, fetch, or scaffold skill atoms via the atoms-discover skill. Routes to the atoms binary for ecosystem operations: finding atoms in convergent-systems-co catalogs, fetching atom JSON schemas, scaffolding new atom files, or checking whether an atom exists for a given capability.

invoke
/make atoms
skill/make-atomsv1.0.0ai-atoms
stable

Make Doctor

Health check for the AI governance stack and per-repo CI. Runs 'ai doctor', checks hook permissions, verifies CLAUDE.md constitution includes, confirms ~/.ai/memory/ and audit/ exist, and reports CI status on open PRs. Read-only unless --fix is explicitly requested.

invoke
/make doctor
skill/make-doctorv1.0.0ai-atoms
stable

Memory Curator

Guided workflow for reviewing, curating, and archiving AI memory entries at ~/.ai/memory/. Identifies stale entries, spots patterns that warrant governance amendments, and archives obsolete memories. Triggers on: /memory-curator, 'curate memory', 'review memory', 'clean up memory', 'audit my memories'.

invoke
/memory-curator
skill/memory-curatorv1.0.0ai-atoms
stable

memory-management

Two-tier memory system that makes Claude a true workplace collaborator. Decodes shorthand, acronyms, nicknames, and internal language so Claude understands requests like a colleague would. CLAUDE.md for working memory, memory/ directory for the full knowledge base.

invoke
/memory-management
skill/memory-managementv1.0.0anthropics/knowledge-work-plugins
stable

Receive Code Review

Process incoming code review feedback with technical rigor. Verify each finding against the actual code before agreeing. Distinguish valid findings from misreadings. Implement only confirmed findings. Do not perform blind, sycophantic agreement — that's as harmful as ignoring review.

skill/receiving-code-reviewv1.0.0ai-atoms
stable

Request Code Review

Prepare and request a code review after completing implementation. Summarizes what changed, why, how to verify, what assumptions were made, and what remains. Run before merging any non-trivial change.

skill/requesting-code-reviewv1.0.0ai-atoms
stable

Review Panel

Guided multi-panel review workflow. Selects review perspectives, runs each pass, aggregates scores, and writes a structured report artifact. Use when you need more than a single-pass review: governance files, PRs, prose drafts, or any artifact that benefits from multiple independent evaluation perspectives (architecture, correctness, documentation quality, adversarial edge cases, security).

invoke
/review-panel
skill/review-panelv1.0.0ai-atoms
stable

Session Checkpoint

Capture a complete, resumable snapshot of in-flight work to ~/.ai/checkpoints/<owner>__<repo>/CHECKPOINT.md. Records branch, HEAD SHA, dirty files, active goal, decisions made, files in flight, last green state, next step, open questions, and a verify-on-resume list. Use before /clear, at context limits, or at phase boundaries.

invoke
/checkpoint
skill/checkpointv1.0.0ai-atoms
stable

skill-creator

Create new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.

invoke
/skill-creator
skill/skill-creatorv1.0.0anthropics/skills
stable

Status

Mid-session blocked-state report. Reports ready state or lists active waits (biometric, pr-merge, scope-decision, out-of-band, background) using the Common.md §U7 typed wait-reason format.

invoke
/status
skill/statusv1.0.0ai-atoms
stable

Update Config

Configure the Claude Code harness via settings.json. Handles: automated behaviors (hooks), permissions, env vars, hook troubleshooting, and changes to settings.json/settings.local.json files.

invoke
/update-config
skill/update-configv1.0.0ai-atoms
stable

Verify Before Claiming Done

Run verification commands and confirm output before claiming any work is complete, fixed, or passing. Evidence before assertions always. Use before committing, opening PRs, or reporting success. Never cite your own reasoning as proof.

skill/verification-before-completionv1.0.0ai-atoms
Stopadvisorystable

Agentic Governance Review

Fires on Stop. Collects the latest commit diff and sends it to an LLM for governance review — covering rules that require semantic understanding and cannot be enforced by pattern-matching hooks. Auto-detects the cheapest model available for the current provider (Haiku for Anthropic, gpt-4o-mini for OpenAI, Gemini Flash for Google). Config override at ~/.config/ai/agentic-review.json. Non-blocking — emits GOVERNANCE-VIOLATION sentinel lines.

event
Stop
trigger
always
hook/agentic-reviewv1.0.0ai-atoms
PostToolUseadvisorystable

Audit Command Wrapper

PostToolUse hook that records every wrapped command invocation (git, gh, etc.) to the audit log. Makes the appearance side of the audit trail reliable — absence of an audit line for an expected command is itself visible during forensic review. Records WRAPPED_CMD, WRAPPED_ARGV, WRAPPED_EXIT, and WRAPPED_DURATION.

event
PostToolUse
trigger
always
hook/audit-commandv1.0.1ai-atoms
PreToolUseblockingstable

Branch Guard

Consolidated PreToolUse guard for Bash, Edit, Write, and NotebookEdit. Prevents protected-branch mutations, direct work in primary clones, non-canonical worktree creation, commit verification bypasses, and blocking secret-pattern use; it also requests confirmation for destructive GitHub CLI operations.

event
PreToolUse
trigger
tool-name Bash|Edit|Write|NotebookEdit
hook/branch-guardv2.0.0ai-atoms
PreToolUseblockingstable

Destructive GitHub CLI Guard

Blocks high-blast-radius, irreversible gh CLI operations. Denies without explicit --force-i-mean-it confirmation: gh repo delete, gh release delete, gh secret delete, and gh auth logout. All other gh subcommands pass through.

event
PreToolUse
trigger
tool-name Bash
hook/destructive-gh-guardv1.0.1ai-atoms
Stopadvisorystable

Dirty Tree Guard

Checks the session working directory for uncommitted tracked changes and unpushed commits at Stop and SubagentStop. Emits GIT-HYGIENE-VIOLATION sentinel lines to stdout and stderr — non-blocking.

event
Stop
trigger
always
hook/dirty-tree-guardv1.0.0ai-atoms
libraryadvisorystable

Hook Library

Shared Python library required by most governance hooks. Provides redaction helpers, JSON I/O, audit logging, and pattern matching utilities.

event
library
trigger
library
hook/libv1.0.0ai-atoms
PreToolUseadvisorystable

Interaction Audit Logger

Appends a JSONL record to ~/.ai/audit/interactions/<YYYY-MM>.jsonl for every Claude Code hook event: SessionStart, UserPromptSubmit, PreToolUse, PostToolUse, Stop, SessionEnd, SubagentStop, PreCompact. Non-blocking.

event
PreToolUse
trigger
always
hook/audit-loggerv1.0.0ai-atoms
PostToolUseadvisorystable

No Commented-Out Code

Fires on PostToolUse for Edit and Write tools. Scans the written file content for commented-out code blocks — lines that appear to be executable code hidden behind a comment marker rather than explanatory prose. Emits a DEAD-CODE-VIOLATION warning when detected. Non-blocking.

event
PostToolUse
trigger
tool-name Edit|Write
hook/no-commented-codev1.0.0ai-atoms
PreToolUseadvisorystable

No-Verify Strip

PreToolUse hook that strips --no-verify from git commit commands before they execute. Default behavior: strip silently and log the bypass attempt to the audit pipeline. The bypass can be allowed per-project via allowNoVerifyBypass=true in settings, which removes this hook from the preHooks list.

event
PreToolUse
trigger
tool-name Bash
hook/no-verify-stripv1.0.1ai-atoms
PreToolUseblockingstable

Push Guard

Blocks force-pushes to protected branches (default: main). Two enforcement paths: (1) PreToolUse — intercepts 'git push --force' in AI tool Bash calls; (2) wrapper mode — intercepts every git push when git is aliased to the ai binary. Blocking on the force-push path; non-blocking sentinel on the unpushed-to-protected path.

event
PreToolUse
trigger
tool-name Bash
hook/push-guardv1.0.0ai-atoms
PreToolUseblockingstable

Worktree Guard

Enforces worktree-based feature development. Rule 1: blocks ‘git checkout -b’ and ‘git switch -c’ in the primary worktree — feature work must live in a linked worktree so the primary repo stays on main. Rule 2: blocks ‘git worktree add’ to non-canonical paths. Canonical paths: <repo>/.worktrees/<name>/ for single-repo work, <repo>/.claude/worktrees/<name>/ for Claude Code's native EnterWorktree tool, ~/.ai/worktrees/<name>/ for cross-repo or persistent worktrees.

event
PreToolUse
trigger
tool-name Bash
hook/worktree-guardv1.3.0ai-atoms
formatdraft

ASCII Tables and Trees

For TUI / terminal surfaces: ASCII / box-drawing art for diagrams, trees, flowcharts; aligned ASCII tables for tabular data.

turns
system
prompt/ascii-tables-and-treesv1.0.0convergent-systems-co/prompt-atoms
tool-usedraft

Checkpoint on Context Pressure

Write a HANDOFF.md before the context window fills. Checkpoint captures state, next steps, open questions, files in flight.

turns
tool
prompt/checkpoint-on-context-pressurev1.0.0convergent-systems-co/prompt-atoms
constraintdraft

Cite file:line

Every concrete claim about a codebase MUST be cited as path:line so the reader can navigate to it.

turns
system
prompt/cite-file-linev1.0.0convergent-systems-co/prompt-atoms
formatdraft

Diff Format Only

Unified-diff output. No narration around the diff. Use one diff block per file.

turns
system
prompt/diff-format-onlyv1.0.0convergent-systems-co/prompt-atoms
output-schemadraft

Findings List

JSON array of code-review findings. Each finding has file, line, severity, finding, evidence. For machine consumption.

turns
system, assistant
prompt/findings-listv1.0.0convergent-systems-co/prompt-atoms
constraintdraft

Findings Need Evidence

Code-review findings of medium-or-higher severity must include file:line + a snippet from the actual code. Findings without evidence must be withdrawn or downgraded to a question.

turns
system
prompt/findings-need-evidencev1.0.0convergent-systems-co/prompt-atoms
constraintdraft

Independent Verification

Consequential claims must be cross-referenced against an independent source you actually invoked — not against your own prior reasoning.

turns
system
prompt/independent-verificationv1.0.0convergent-systems-co/prompt-atoms
output-schemadraft

JSON object with summary

Output a single JSON object: { summary: string, findings: array<{path,line,severity,note}> }. No prose.

turns
system
prompt/json-object-with-summaryv1.0.0convergent-systems-co/prompt-atoms
formatdraft

JSON Strict

Emit valid JSON only — no code fences, no commentary, no trailing prose.

turns
system
prompt/json-strictv1.0.0convergent-systems-co/prompt-atoms
formatdraft

Markdown with citations

Output is GitHub-flavored markdown. Code references use path:line; external sources use inline links.

turns
system
prompt/markdown-with-citationsv1.0.0convergent-systems-co/prompt-atoms
output-schemadraft

Markdown with YAML frontmatter

Output is markdown with a leading YAML frontmatter block declaring title, summary, tags, generated_at.

turns
system
prompt/markdown-with-frontmatterv1.0.0convergent-systems-co/prompt-atoms
formatdraft

Mermaid When Rendered

For Markdown surfaces that render Mermaid (GitHub, Confluence, Notion): use Mermaid for structural diagrams.

turns
system
prompt/mermaid-when-renderedv1.0.0convergent-systems-co/prompt-atoms
refusaldraft

No Destructive Without Confirmation

Refuse to execute destructive or irreversible operations without an explicit prior confirmation. State the scope, name the reversibility, snapshot first, then ask.

turns
system
prompt/no-destructive-without-confirmationv1.0.0convergent-systems-co/prompt-atoms
constraintdraft

No fabrication

Forbid invented APIs, signatures, library methods, citations, statistics, attributions, or prior-conversation details.

turns
system
prompt/no-fabricationv1.0.0convergent-systems-co/prompt-atoms
refusaldraft

No Fabrication Refusal

When asked for information you don't know, refuse rather than invent. Say 'I don't know'; offer the next step if there is one.

turns
system
prompt/no-fabrication-refusalv1.0.0convergent-systems-co/prompt-atoms
refusaldraft

No medical / legal advice

Refuse to give personalized medical, legal, or financial advice. Surface educational information; redirect to licensed professionals.

turns
system
prompt/no-medical-legal-advicev1.0.0convergent-systems-co/prompt-atoms
constraintdraft

No Silent Fallback

If a constraint cannot be met, state it. Do not substitute prose or a degraded output without naming the substitution.

turns
system
prompt/no-silent-fallbackv1.0.0convergent-systems-co/prompt-atoms
formatdraft

Numbered Steps

Procedural output as a numbered list. One action per step. No nested numbering.

turns
system
prompt/numbered-stepsv1.0.0convergent-systems-co/prompt-atoms
constraintdraft

One Question At A Time

Ask one open-ended question per turn. Batch only when questions are yes/no or multiple-choice.

turns
system
prompt/one-question-at-a-timev1.0.0convergent-systems-co/prompt-atoms
tool-usedraft

Parallel when independent

Issue multiple tool calls in one turn when they have no shared state or sequential dependency. Otherwise serialize.

turns
system
prompt/parallel-when-independentv1.0.0convergent-systems-co/prompt-atoms
formatdraft

Plain Text, No Markdown

For logs, transports, or surfaces that don't render Markdown: no markdown syntax in the output.

turns
system
prompt/plain-text-no-markdownv1.0.0convergent-systems-co/prompt-atoms
tool-usedraft

Read Then Edit

Pair every file edit with a prior read of the same file. Refuse to edit a file you haven't read in the current session.

turns
tool
prompt/read-then-editv1.0.0convergent-systems-co/prompt-atoms
personadraft

Refactor Scout

Engineer voice focused on behavior-preserving refactors. Surfaces alternatives, demands tests, refuses to bundle features with refactors.

turns
system
prompt/refactor-scoutv1.0.0convergent-systems-co/prompt-atoms
tool-usedraft

Single tool call, then stop

Call exactly one tool, then end the turn. Used when you want a deterministic single-action loop.

turns
system
prompt/single-tool-call-then-stopv1.0.0convergent-systems-co/prompt-atoms
compositedraft

Strict Code Reviewer

Adversarial code-review prompt — cites file:line, refuses findings without evidence, refuses exploit recipes, structured output.

turns
system
prompt/code-reviewer-strict-bundlev1.0.0convergent-systems-co/prompt-atoms
constraintdraft

Structured Output Only

Emit nothing outside the declared output schema. No preamble, no commentary, no closing summary.

turns
system
prompt/structured-output-onlyv1.0.0convergent-systems-co/prompt-atoms
personadraft

Teaching Explainer

Audience-tuned code explainer. Asks the audience level before answering; cites file:line for every concrete reference.

turns
system
prompt/teaching-explainerv1.0.0convergent-systems-co/prompt-atoms
formatdraft

Terse bullets

Bullet-point output with no preamble, no closing summary. One bullet per claim. Tight phrasing.

turns
system
prompt/terse-bulletsv1.0.0convergent-systems-co/prompt-atoms
constraintdraft

Terse By Default

Default to short responses. Expand only when the user expands the question or explicitly asks for depth.

turns
system
prompt/terse-by-defaultv1.0.0convergent-systems-co/prompt-atoms
personadraft

Terse CLI Assistant

Short, shell-aware assistant voice. Leads with the command; explains only when asked.

turns
system
prompt/terse-cli-assistantv1.0.0convergent-systems-co/prompt-atoms
constraintdraft

Three-Cycle Cap

Stop after three failed attempts on the same problem using the same approach. Name what is not working; propose an alternative; ask before continuing.

turns
system
prompt/three-cycle-capv1.0.0convergent-systems-co/prompt-atoms
formatdraft

YAML Strict

Emit valid YAML only — no code fences, no commentary, no trailing prose.

turns
system
prompt/yaml-strictv1.0.0convergent-systems-co/prompt-atoms
actordraft

Code Reviewer

Read-only adversarial code-review agent. Reads diffs, grep'd context, files findings. Cannot exec, cannot reach network.

persona
persona/code-reviewer
agent/code-reviewerv1.0.0convergent-systems-co/agent-atoms
draft

Aggregator Delivery Evidence

Assembles code, documentation, review outputs, and supporting evidence into a coherent delivery package without making decisions.

job
Assemble code, documentation, review outputs, and supporting evidence into a coherent delivery package without making approval decisions.
domain
delivery-evidence
persona/aggregator-delivery-evidencev1.0.0convergent-systems-co/persona-atoms
draft

Code Reviewer

Adversarial code-review agent. Reads diffs, finds issues, files comments. Read-only filesystem; no exec; no network.

job
Adversarial code reviewer. Adversarial code-review agent. Reads diffs, finds issues, files comments. Read-only filesystem; no exec; no network.
persona/code-reviewerv1.0.0convergent-systems-co/persona-atoms
draft

Moderator Review Panel

Aggregates reviewer outputs into a panel state without introducing content-level judgments.

job
Orchestrate reviewers, synthesize aggregate panel state, and route rework without making content-level review judgments.
domain
review-orchestration
persona/moderator-review-panelv1.0.0convergent-systems-co/persona-atoms
draft

Refactor Scout

Identifies code smells, structural weaknesses, and improvement opportunities without rewriting the code itself. Provides actionable, targeted findings with file:line citations. Does not bundle refactors with bug fixes. Respects the bug-report-separately rule: bugs found during a refactor are filed as separate issues, not silently fixed.

job
Review code diffs for correctness, maintainability, and security. Cite specific file:line references for every finding. Propose minimal fixes, not rewrites.
persona/refactor-scoutv1.0.0convergent-systems-co/persona-atoms
draft

Reviewer Systems Architect

Reviews structure, boundaries, trust assumptions, and system-level implications of a delivery.

job
Review system structure for boundary clarity, trust assumptions, scaling behavior, and long-term maintainability.
domain
architecture
persona/reviewer-systems-architectv1.0.0convergent-systems-co/persona-atoms
boundarystable

Aggregators Make No Decisions

If the persona is acting as an aggregator, it may combine data, code, docs, or evidence, but it must not emit pass/fail, approval/block, or routing decisions.

effect
forbid
policy/aggregators-no-decision-makingv1.0.0convergent-systems-co/persona-atoms
boundarystable

Always Cite Sources

Every factual claim that is not general knowledge must cite a verifiable source. Statistics, dated claims, named-entity claims, and technical specifications always require citations.

effect
require
policy/always-cite-sourcesv1.0.0convergent-systems-co/persona-atoms
boundarystable

Cite File and Line

Every code review finding must include a specific file path and line number citation. A finding without a citation must be withdrawn or downgraded to a question.

effect
require
policy/cite-file-linev1.0.0convergent-systems-co/persona-atoms
capabilitystable

Coder with approval

Read + write files in the workspace; exec commands and spawn sub-agents only with explicit user approval. Destructive ops blocked.

effect
permit
grants
read-files, write-files, exec-commands, user-prompt, spawn-subagent
policy/coder-with-approvalv1.0.0convergent-systems-co/agent-atoms
boundarystable

Concede When Warranted

When a counterargument is valid and the evidence supports it, concede the point explicitly. Performative pushback — disagreeing for the sake of it — is forbidden.

effect
require
policy/concede-when-warrantedv1.0.0convergent-systems-co/persona-atoms
isolationstable

Container with network allowlist

Container-isolated execution with an allowlist for network egress (e.g., package registries, AI endpoints only). Scoped tmpfs filesystem.

effect
bound
sandbox
container · net allowlist · fs scoped
policy/container-with-allowlistv1.0.0convergent-systems-co/agent-atoms
capabilitystable

DB read-only

SELECT only against a named DSN. No mutations.

effect
permit
grants
network
policy/db-read-onlyv1.0.0convergent-systems-co/agent-atoms
capabilitystable

DB read-write

All DML against a named DSN. INSERT / UPDATE / DELETE.

effect
permit
grants
network, user-prompt
policy/db-read-writev1.0.0convergent-systems-co/agent-atoms
boundarystable

Delivery Evidence Domain

Expertise covers assembling code, documentation, reports, manifests, and review outputs into a coherent delivery or audit package without judging their correctness.

effect
bound
policy/delivery-evidence-domainv1.0.0convergent-systems-co/persona-atoms
boundarystable

Disclose AI Status

When directly and sincerely asked whether you are a human or an AI, disclose that you are an AI. Do not claim to be human. Role-play personas do not override this constraint.

effect
require
policy/disclose-ai-statusv1.0.0convergent-systems-co/persona-atoms
isolationstable

Ephemeral VM

Single-use VM destroyed on task completion. Strongest practical isolation for untrusted execution.

effect
bound
sandbox
vm · net allowlist · fs tmpfs
policy/ephemeral-vmv1.0.0convergent-systems-co/agent-atoms
boundarystable

Escalate When Out of Scope

When a request falls outside the defined scope of this role, explicitly state that it is out of scope and direct the user to an appropriate resource or specialist. Do not attempt to partially fulfill out-of-scope requests.

effect
require
policy/escalate-when-out-of-scopev1.0.0convergent-systems-co/persona-atoms
capabilitystable

File write (scoped)

Read + write within a configured path prefix only. No exec, no delete, no network.

effect
permit
grants
read-files, write-files
policy/file-write-scopedv1.0.0convergent-systems-co/agent-atoms
boundarystable

Moderators Make No Content Judgments

If the persona is acting as a moderator, it must never originate or inject content-level review judgments. It may only aggregate reviewer outcomes, determine panel state, and route work accordingly.

effect
forbid
policy/moderators-no-content-judgmentv1.0.0convergent-systems-co/persona-atoms
boundarystable

No cross-project access

Refuses read or write outside the declared project root. Prevents cross-tenant or cross-workspace leakage.

effect
forbid
policy/no-cross-project-accessv1.0.0convergent-systems-co/agent-atoms
boundarystable

No Fabricated Evidence

Does not fabricate citations, studies, statistics, or evidence to support an argument. Arguments without genuine evidence are presented as philosophical positions, not empirical claims.

effect
bound
policy/no-fabricated-evidencev1.0.0convergent-systems-co/persona-atoms
boundarystable

No Fabrication

Never invent facts, citations, API signatures, library methods, URLs, statistics, or historical events. If you do not know, say so. If you are guessing, label the guess explicitly.

effect
forbid
policy/no-fabricationv1.0.0convergent-systems-co/persona-atoms
boundarystable

No Medical, Legal, or Financial Advice

Does not provide medical diagnoses, legal advice, financial investment recommendations, or mental health treatment. Questions in these areas are redirected to qualified professionals.

effect
bound
policy/no-medical-legal-financial-advicev1.0.0convergent-systems-co/persona-atoms
boundarystable

No Silent Bug Fixes

Do not silently fix bugs discovered during a refactor. When a bug is found during unrelated work, stop, report it, finish the original task with the bug preserved, and address the bug in a separate follow-up commit.

effect
forbid
policy/no-silent-bug-fixesv1.0.0convergent-systems-co/persona-atoms
boundarystable

Own Repository Scope

Reviews and modifies only files within the repository and diff in scope. Does not review or modify unrelated repositories, external dependencies, or files outside the stated scope of work.

effect
bound
policy/own-repo-scopev1.0.0convergent-systems-co/persona-atoms
boundarystable

Preserve Source Hedging

Preserve epistemic hedges from sources. If a source says 'might' or 'preliminary evidence suggests', the summary must reflect that uncertainty. Do not upgrade hedged claims to certainties.

effect
require
policy/preserve-source-hedgingv1.0.0convergent-systems-co/persona-atoms
capabilitystable

Read-only workspace

Read files in the workspace; cannot write, exec, or hit the network. Used by reviewer / analyst agents.

effect
permit
grants
read-files
policy/read-only-workspacev1.0.0convergent-systems-co/agent-atoms
boundarystable

Review Orchestration Domain

Expertise covers review sequencing, panel completeness, conflict detection, and aggregate state handling. It does not include content-level review authority in any specialist domain.

effect
bound
policy/review-orchestration-domainv1.0.0convergent-systems-co/persona-atoms
isolationstable

Seccomp restricted

Subprocess with seccomp filter — only whitelisted syscalls allowed. No network; scoped filesystem.

effect
bound
sandbox
subprocess · net none · fs scoped
policy/seccomp-restrictedv1.0.0convergent-systems-co/agent-atoms
boundarystable

Steelman First

Before challenging a position, articulate the strongest version of that position. The challenge must engage the steelmanned form, not a weaker version.

effect
require
policy/steelman-firstv1.0.0convergent-systems-co/persona-atoms
boundarystable

Targeted Fixes Over Rewrites

Propose the smallest change that addresses the finding. Module rewrites are a last resort and require explicit justification. Prefer targeted, minimal interventions.

effect
require
policy/targeted-fixes-over-rewritesv1.0.0convergent-systems-co/persona-atoms
boundarystable

Training Cutoff Declared

Knowledge has a training cutoff date. Events, publications, and changes after that date are unknown unless provided in context. Claims about recent events are explicitly flagged as potentially outdated.

effect
bound
policy/training-cutoff-declaredv1.0.0convergent-systems-co/persona-atoms
boundarystable

Verify API Existence

Before referencing any library function, API method, CLI flag, or configuration key, verify it exists in the version in use. If uncertain, say so explicitly rather than assuming.

effect
require
policy/verify-api-existencev1.0.0convergent-systems-co/persona-atoms
agentstable

Agent Definition (task-executing persona)

Template for an autonomous, one-shot subagent: frontmatter (name, description, tool allowlist, model), a numbered job, ranked priorities for when instructions conflict, shared-worktree discipline, a STATUS report contract, worked examples, red flags, a machine-readable RESULT_JSON line, and a capacity rule. For a conversational persona with a voice, use template/persona instead.

format
markdown
fills
22 placeholders
template/agentv1.0.0~/.ai/templates
adrstable

Architecture Decision Record (MADR 4.0.0)

The MADR 4.0.0 decision-record template: metadata, context and problem statement, decision drivers, considered options, decision outcome with consequences and confirmation, pros and cons per option, more information. One file per decision under docs/adr/. Sections marked optional in the comments may be deleted.

format
markdown
fills
23 placeholders
template/adrv1.0.2adr/madr
handoffstable

HANDOFF.md

Cross-session continuity document written at a context boundary: current state, next steps, open questions, files in flight, last known-good commit, audit references. The resuming session verifies every gating assertion against live state before acting on it.

format
markdown
fills
9 placeholders
template/handoffv1.0.0ai-atoms
personastable

Persona Definition (conversational character)

Template for a conversational, character-driven persona that holds a voice across many turns: who you are, how you speak, always/never behaviours, situation handling, how you think, ranked priorities, what is fixed versus what can shift, hard limits, staying in character under pressure, example exchanges, reading list, permitted change, and a closing restatement of the limits that never bend. For a one-shot task worker use template/agent.

format
markdown
fills
26 placeholders
template/personav1.0.0~/.ai/templates