Skill Claude CodeCodex
Scaffold a new spec-compliant, portable Agent Skill (folder + SKILL.md with YAML frontmatter where name == the folder, plus optional scripts/ and references/) and validate it against the agentskills.io standard. When asked to "create a skill", "scaffold a skill", "author a new skill", "make a SKILL.md", "turn this…
Skill Claude CodeCodex
Security gate that vets any EXTERNAL skill, MCP server, or plugin BEFORE it is trusted or installed. When asked to "vet a skill", "review this MCP server", "is this plugin safe", "audit a downloaded skill", or before skill add installs anything, scan the SKILL.md / manifest for prompt-injection and…
Skill Claude CodeCodex
Find the root cause of a bug methodically instead of guessing. When asked to "debug this", "why is this failing", "this test is flaky", "track down this crash", "find the root cause", "this worked before and now it doesn't", or when a fix attempt failed and you are tempted to try random changes, run a reproduce →…
Skill Claude CodeCodex
Author and run focused, high-signal tests for code, the TDD way. When asked to "write tests", "add test coverage", "do TDD", "test this function", "cover this edge case", "get this to green", or when a change lands without tests, infer the project's existing test framework, write targeted tests for the changed or…
Skill Claude CodeCodex
Enumerate threats at design time using STRIDE, data-flow diagrams, and trust-boundary analysis. When asked to "threat model this", "do a STRIDE analysis", "what could go wrong with this design", "map the attack surface", "find the trust boundaries", or when scoping a new system, integration, agent loop, or data flow…
Skill Claude CodeCodex
Marking a UI task "complete" because the build and unit tests pass is risky — units don't catch routing bugs, hydration issues, race conditions on real network calls, broken visual states, or wrong API payloads. A real browser hitting the real running app is the only honest "complete" for a UI task. This skill is…
Cursor rule Cursor
You MUST read AGENTS.md before writing code or planning tasks. Start there. For high-level project state, resolve the external knowledge base with scripts/agentware config --knowledge-dir-only and read its MAIN.md.
Instructions file GitHub Copilot
Instructions for r5rana/agentware: You MUST read AGENTS.md before writing code or planning tasks. Start there. For high-level project state, resolve the external knowledge base with scripts/agentware config --knowledge-dir-only and read its MAIN.md.
Instructions file CodexOpenCode
Instructions for r5rana/agentware, covering execution loop, build and complete, planning, knowledge base location and package immutability & self-extension.
Instructions file
Instructions for r5rana/agentware, covering agentware — claude code project instructions, 🔴 bootstrap gate — check first, before any task, canonical methodology (imported, always-on), always-on steering (imported) and working in this repo.