All authors
odylith avatar

Claude Skills by odylith

github.com/odylith
180 skillsA× 1800 installs29 views
Odylith SyncA

Refresh governed Odylith surfaces for the current changed slice.

ai-agentsgo
0
36
Odylith VersionA

Inspect the pinned, active, and locally available Odylith versions.

ai-agents
0
36
Fail Closed Code HygieneA

Use this maintainer-only skill when a pass is explicitly about removing AI slop, fake modularization, duplicate helper churn, mirror drift, oversized-file pressure, or other structural debt in shipped Odylith code. Treat AI slop as a regression. Apply that bar to any codebase or project surface in this lane: services, libraries, apps, CLIs, infra glue, scripts, docs, prompts, hooks, templates, config, and generated assets all count. Structural cleanup is not permission to drift behavior, UX, ...

ai-agentspythongo
0
36
Odylith Github Issue TriageA

Use this maintainer-only skill when the operator gives a GitHub issue URL or number and wants Odylith to fetch it, classify it, map it to Casebook/Radar/Registry truth, and draft the public GitHub response. Do not mirror this skill into consumer-safe shared skills, `.agents/skills`, `.claude/skills`, or bundled install assets.

ai-agentsgobash
0
36
Odylith Github Release CloseoutA

Use this maintainer-only skill when the operator needs to prepare or apply public GitHub issue updates for issues fixed in a release. Do not mirror this skill into consumer-safe shared skills, `.agents/skills`, `.claude/skills`, or bundled install assets.

ai-agentsgobash
0
36
Release Benchmark PublishingA

Use this skill only in the Odylith product repo when preparing a release or refreshing the public benchmark story in the repo-root `README.md`. Do not use this skill in consumer repos.

ai-agentspythongo
0
36
Odylith Atlas Auto UpdateA

Use this skill only when the user explicitly invokes `$odylith-atlas-auto-update` or asks to refresh Atlas diagrams from change-watch metadata. 1. Run `./.odylith/bin/odylith atlas auto-update --repo-root .`. 2. If Atlas auto-update reports no changed paths, say so plainly. 3. If freshness or render gates fail, surface the exact blocker and next command instead of widening the task.

ai-agents
0
36
Odylith Atlas RenderA

Use this skill only when the user explicitly invokes `$odylith-atlas-render` or asks to render Atlas from the Mermaid catalog. 1. Run `./.odylith/bin/odylith atlas render --repo-root .`. 2. Report whether Atlas rendered cleanly or surfaced a freshness blocker. 3. If freshness blocks the render, say which follow-up command is needed rather than retrying blindly.

ai-agents
0
36
Odylith Backlog CreateA

Use this skill only when the user explicitly invokes `$odylith-backlog-create` or asks to create one or more Radar backlog workstreams. 1. Search existing Radar, plans, bugs, and recent Compass context first so you extend the right record instead of duplicating it. 2. Gather grounded core detail before authoring: Problem, Customer, Opportunity, Product View, and Success Metrics. Do not use title-derived boilerplate, `TBD`, `Details.`, or other placeholders. Exception: when the user asks for a...

ai-agents
0
36
Odylith Backlog ValidateA

Use this skill only when the user explicitly invokes `$odylith-backlog-validate` or asks to validate Radar backlog and plan linkage contracts. 1. Run `./.odylith/bin/odylith validate backlog-contract --repo-root .`. 2. Report the validated counts or the failing contract plainly. 3. Treat core-detail failures as real authoring defects: Problem, Customer, Opportunity, Product View, and Success Metrics must be grounded and cannot be placeholder or backlog-create boilerplate. 4. If the validator ...

ai-agentsgo
0
36
Odylith Casebook Bug CaptureA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Casebook Bug InvestigationA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Casebook Bug PreflightA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Code Hygiene GuardA

Use this skill when a slice shows duplicate helper churn, fake modularization, oversized-file pressure, mirrored host drift, or comment slop. Treat AI slop as a regression. This skill is shared across consumer and maintainer lanes, and it applies on both Codex and Claude. Codex and Claude must enforce the same anti-slop contract across consumer and maintainer lanes. Consumer repos may be Python, TypeScript, JavaScript, Go, Rust, Java, shell, SQL, or mixed-language; the language changes, the a...

ai-agentsjavascripttypescript
0
36
Odylith Compass ExecutiveA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Compass LogA

Use this skill only when the user explicitly invokes `$odylith-compass-log` or asks to append a bounded execution note into Compass. 1. Identify the current workstream, entry kind, and one-sentence summary worth preserving. 2. Run `./.odylith/bin/odylith compass log --repo-root . --kind <kind> --summary "<summary>"`. 3. Add `--workstream` and `--component` when the active slice is known and the command needs those anchors explicitly. 4. Keep the log entry factual, short, and specific to the c...

ai-agents
0
36
Odylith Compass RefreshA

Use this skill only when the user explicitly invokes `$odylith-compass-refresh` or asks to refresh Compass and wait for the current pass to settle. 1. Run `./.odylith/bin/odylith compass deep-refresh --repo-root .`. 2. Report whether the refresh completed, deferred, or surfaced a provider or runtime blocker. 3. If the refresh is blocked, summarize the blocker instead of retrying blindly.

ai-agents
0
36
Odylith Compass Timeline StreamA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Component RegistryA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Context Engine OperationsA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsrustgo
0
36
Odylith ContextA

Use this skill only when the user explicitly invokes `$odylith-context` or asks to resolve a known Odylith anchor into the smallest useful local context. 1. Use this only after the current turn has run `odylith start`, or after a fresh session-start packet already exists and an exact anchor is known, such as a workstream id, bug id, component, diagram id, or repo path. 2. Do not run this in parallel with `odylith start`; startup is the serial gate and context is the follow-on exact-anchor pac...

ai-agentsgo
0
36
Odylith Delivery Governance Surface OpsA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsrustgo
0
36
Odylith Diagram CatalogA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith DisciplineA

Use this skill when Odylith Discipline behavior, credit-safe hard laws, learning-spine signals, or benchmark-proved pressure behavior need validation or implementation.

ai-agentsgoshell
0
36
Odylith DoctorA

Use this skill only when the user explicitly invokes `$odylith-doctor` or asks to verify or repair the current Odylith install. 1. Run `./.odylith/bin/odylith doctor --repo-root .` for inspection. 2. Add `--repair` only when the user explicitly asks for repair or when the current task is clearly repair-authorized. 3. Add `--reset-local-state` only when the user explicitly wants a stronger local-state reset. 4. Report the verified posture or blocker instead of retrying blindly.

ai-agents
0
36
Odylith Greenfield GovernanceA

Use this skill when the operator asks Odylith to build, govern, plan, or architect a new project before source code exists. 1. Do not refuse merely because the repo has no app source. Greenfield intent is valid proposal evidence, not source evidence. 2. Run the host-reasoning contract path: `./.odylith/bin/odylith greenfield propose --repo-root . --prompt "<operator request>"`. 3. Use the active host model's full reasoning to draft the proposal in normal chat. The CLI output is the evidence/s...

ai-agentsgo
0
36
Odylith Guidance BehaviorA

Use this skill when guidance behavior pressure cases, high-risk agent instructions, or cross-host guidance contracts need deterministic validation. Guidance Behavior is the first deterministic pressure-family proof lane under Odylith Discipline; it proves concrete hard-law families while the broader Odylith Discipline layer handles open-world stance, credit budgets, learning signals, and benchmark sovereignty.

ai-agentsgo
0
36
Odylith QueryA

Use this skill only when the user explicitly invokes `$odylith-query` or asks to search the local Odylith projection store after concrete anchors already exist. 1. Confirm that a concrete noun already exists, such as a workstream, component, path, bug id, or diagram id. 2. Run `./.odylith/bin/odylith query --repo-root . "<terms>"`. 3. Use the result to narrow the next file reads, tests, or edits instead of expanding into broad unguided repo search. 4. Do not use this as a substitute for the i...

ai-agents
0
36
Odylith Registry Spec SyncA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Registry Sync SpecsA

Use this skill only when the user explicitly invokes `$odylith-registry-sync-specs` or asks to sync mapped Compass requirement evidence into component living specs. 1. Run `./.odylith/bin/odylith governance sync-component-spec-requirements --repo-root .`. 2. Report which components or forensics records updated. 3. If nothing changed, say that plainly instead of implying hidden work.

ai-agentsgo
0
36
Odylith Registry ValidateA

Use this skill only when the user explicitly invokes `$odylith-registry-validate` or asks to validate Registry component inventory contracts. 1. Run `./.odylith/bin/odylith validate component-registry --repo-root .`. 2. Report the contract result, component counts, and any critical policy finding. 3. Keep follow-up bounded to the surfaced Registry contract issue.

ai-agents
0
36
Odylith Release PlanningA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsrustgo
0
36
Odylith Schema Registry GovernanceA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Security HardeningA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentspythonrust
0
36
Odylith Session BriefA

Use this skill only when the user explicitly invokes `$odylith-session-brief` or asks for a concise session brief grounded in current Odylith state. 1. Run `./.odylith/bin/odylith session-brief --repo-root .`. 2. Use current workstream, Compass, Casebook, and validation evidence that already exists. 3. Summarize what moved, what is blocked, and the next concrete move. 4. Prefer grounded evidence over narration flourishes.

ai-agents
0
36
Odylith Session ContextA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Show MeA

Use this skill when the operator says "show me what you can do", "what can Odylith do", "what can you do for this repo", or any similar first-time exploration request. Works identically on Claude Code and Codex. Do not use this skill for `Odylith, help`. That request should route directly to the CLI help output and print stdout only. This is a first-match route lock. If you have not run a show command and captured stdout, do not answer. Never replace `odylith show` stdout with a hand-written ...

ai-agentsrustgo
0
36
Odylith StartA

Use this skill only when the user explicitly invokes `$odylith-start` or directly asks to run the repo-local Odylith startup contract. 1. Run `./.odylith/bin/odylith start --repo-root .` for substantive work. Do not run `odylith context`, `odylith query`, `git status`, or broad repo search in parallel with this step; let startup finish first so the host transcript reflects the real grounding order. 2. If startup cannot narrow the slice but the user already named a workstream, bug, component, ...

ai-agentsgit
0
36
Odylith Subagent OrchestratorA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgo
0
36
Odylith Subagent RouterA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgoapi
0
36
Odylith SyncA

Use this skill only when the user explicitly invokes `$odylith-sync` or asks to refresh governed Odylith surfaces for the current changed slice. 1. Identify the changed source-of-truth paths under `odylith/radar/source/`, `odylith/technical-plans/`, `odylith/casebook/bugs/`, `odylith/registry/source/`, `odylith/atlas/source/`, or other governed Odylith paths. 2. Run `./.odylith/bin/odylith sync --repo-root . --impact-mode selective <changed_paths...>`. 3. If the change only needs a narrow rer...

ai-agentsgo
0
36
Odylith VersionA

Use this skill only when the user explicitly invokes `$odylith-version` or asks to inspect the pinned, active, and locally available Odylith versions. 1. Run `./.odylith/bin/odylith version --repo-root .`. 2. Report the active pinned runtime posture plainly. 3. When relevant, distinguish the active target release from the latest shipped release instead of collapsing them together.

ai-agents
0
36
Odylith Atlas Auto UpdateA

Use this skill only when the user explicitly invokes `$odylith-atlas-auto-update` or asks to refresh Atlas diagrams from change-watch metadata. 1. Run `./.odylith/bin/odylith atlas auto-update --repo-root .`. 2. If Atlas auto-update reports no changed paths, say so plainly. 3. If freshness or render gates fail, surface the exact blocker and next command instead of widening the task.

ai-agents
0
36
Odylith Atlas RenderA

Use this skill only when the user explicitly invokes `$odylith-atlas-render` or asks to render Atlas from the Mermaid catalog. 1. Run `./.odylith/bin/odylith atlas render --repo-root .`. 2. Report whether Atlas rendered cleanly or surfaced a freshness blocker. 3. If freshness blocks the render, say which follow-up command is needed rather than retrying blindly.

ai-agents
0
36
Odylith Backlog CreateA

Use this skill only when the user explicitly invokes `$odylith-backlog-create` or asks to create one or more Radar backlog workstreams. 1. Search existing Radar, plans, bugs, and recent Compass context first so you extend the right record instead of duplicating it. 2. Gather grounded core detail before authoring: Problem, Customer, Opportunity, Product View, and Success Metrics. Do not use title-derived boilerplate, `TBD`, `Details.`, or other placeholders. Exception: when the user asks for a...

ai-agents
0
36
Odylith Backlog ValidateA

Use this skill only when the user explicitly invokes `$odylith-backlog-validate` or asks to validate Radar backlog and plan linkage contracts. 1. Run `./.odylith/bin/odylith validate backlog-contract --repo-root .`. 2. Report the validated counts or the failing contract plainly. 3. Treat core-detail failures as real authoring defects: Problem, Customer, Opportunity, Product View, and Success Metrics must be grounded and cannot be placeholder or backlog-create boilerplate. 4. If the validator ...

ai-agentsgo
0
36
Odylith Casebook Bug CaptureA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Casebook Bug InvestigationA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Casebook Bug PreflightA

- CLI-first is non-negotiable for both Codex and Claude Code. Remove all hand-authoring for places where Odylith CLI should be doing the heavy-lifting. When an Odylith CLI command exists for an operation, you must call the CLI command and you must not hand-edit governed files the CLI owns. Hand-authoring governed truth where a CLI exists is a hard policy violation, not a stylistic preference. The authoritative policy, CLI surface enumeration, allowed hand-edit surfaces, and failure-mode handling

ai-agentsgobash
0
36
Odylith Code Hygiene GuardA

Use this skill when a slice shows duplicate helper churn, fake modularization, oversized-file pressure, mirrored host drift, or comment slop. Treat AI slop as a regression. This skill is shared across consumer and maintainer lanes, and it applies on both Codex and Claude. Codex and Claude must enforce the same anti-slop contract across consumer and maintainer lanes. Consumer repos may be Python, TypeScript, JavaScript, Go, Rust, Java, shell, SQL, or mixed-language; the language changes, the a...

ai-agentsjavascripttypescript
0
36