Skill Claude CodeCodex
A repository-local process for importing selected third-party skills into a CSL Agent Kit codebase while recording where each skill came from. It preserves source details such as the repository, path, version reference, commit, licence, and upstream status.
Plugin Claude Code
Personal agent toolkit by CSL.
Plugin Claude Code
CSL Agent Kit: skills, commands, hooks, and workflow tools for Claude Code, Codex, Cursor, and Pi.
Instructions file CodexOpenCode
Instructions for SSBun/csl-agent-kit, covering project agent rules, rule changes and task files.
Command
Systematically update project documentation with implementation status, API changes, and synchronized content across specs and agent context files.
Command
Activate a built-in SOP as a Claude Code rule. Cursor users should use agent-sops instead.
Skill Claude CodeCodex
Analyze one Git project or component and write one systematic, source-backed current-state report under docs/analysis/. Use for durable project maps, component maps, responsibility and functional-module models, core working flows, or architecture understanding needed for continued development. Exclude quick…
Skill Claude CodeCodex
Use when the user wants to release a project. This skill only routes to the matching release SOP, checks readiness, and gathers confirmation items; it does not perform ecosystem-specific publishing itself.
Skill Claude CodeCodex
A Chinese-language skill for searching Zhihu, a Chinese question-and-answer community, and the wider web, plus reading the current user's Zhihu content.
Skill Claude CodeCodex
Maintain this repository's project-local Task Target alignment evaluations. Use in CSL Agent Kit for level and Safety cases, fresh-context model runs, scoring, and protocol comparisons.
Hook
Runs when a session starts on startup, resume, clear, compact and new, executing sop-summaries.sh and agent-hooks.js via sh with --show-toplevel (2 commands). From SSBun/csl-agent-kit.
Hook
Runs after the context is compacted for manual and auto tool calls, executing sop-summaries.sh and agent-hooks.js via sh with --show-toplevel (2 commands). From SSBun/csl-agent-kit.
Hook
Runs when you submit a prompt, before the agent sees it, executing sop-candidates.js and agent-hooks.js via sh with --show-toplevel (2 commands). From SSBun/csl-agent-kit.
Hook
Runs before the agent uses a tool for Bash, Edit, Write and apply_patch tool calls, executing agent-hooks.js via sh (2 commands). From SSBun/csl-agent-kit.
Hook
Runs after a tool call finishes for mcp__.*figma.*(get_design_context, get_metadata, get_screenshot, get_figjam, get_variable_defs, get_libraries, search_design_system), mcp__mastergo_magic_mcp.*(getDesignSections, getDsl, getD2c, getMeta, getDesignTexts, getDesignSvgs and extractSvg) tool calls, executing…
Hook
Runs when the agent asks permission for an action, executing agent-hooks.js via sh. From SSBun/csl-agent-kit.
Hook
Runs before the context is compacted for manual and auto tool calls, executing agent-hooks.js via sh. From SSBun/csl-agent-kit.
Hook
Runs when a subagent starts, executing agent-hooks.js via sh. From SSBun/csl-agent-kit.
Hook
Runs when a subagent finishes, executing agent-hooks.js via sh. From SSBun/csl-agent-kit.
Hook
Runs when the agent finishes a response, executing agent-hooks.js via sh. From SSBun/csl-agent-kit.
Agent
General-purpose subagent carrier. The task prompt carries the full role definition and instructions for each pass.
Skill Claude CodeCodex
Analyze one Git project, directory, or file against one current implementation question and write one durable, source-backed report under docs/analysis/. Use for /analyze-project, need-bounded deep source analysis, a current behavior chain, or repository-specific prediction checks. Develop and learn are focus aliases…
Skill Claude CodeCodex
Run a fail-closed, uncapped Reviewer–Editor loop for code, PRDs, RFCs, design documents, and other deliverables. Use only when the user explicitly requests adversarial review, a two-agent Reviewer–Editor loop, or independent Reviewer approval. Exclude ordinary one-pass review, self-review, routine file operations, and…
Agent
Audits and remediates findings from the adversarial-reviewer. Judges adequacy, minimal resolution, blast radius, and proportionality before accepting, narrowing, rejecting, acknowledging, or escalating each finding. Implements the smallest sufficient fix and never self-approves. Used by the adversarial-review skill as…