Rune

Performs adversarial red-team analysis on approved plans to identify edge cases, security risks, scalability issues, error paths, and integration risks befor...

What to check before installing: - Confirm source/trust: The SKILL.md points to a GitHub repo and docs site, but the registry entry lists 'Source: unknown' and no homepage. Verify the upstream repository and maintainer identity (open the repo URL, inspect commit history, stars, and releases). Do not install unless you trust the publisher. - Inspect code files: The bundle includes src/index.ts and multiple scripts. If possible, open and grep for network calls, hardcoded endpoints, command execution (child_process/exec), or references to tokens. Search for 'curl', 'gh api', 'fetch(', 'axios', 'exec(', 'spawn(', 'process.env.' and review what would be transmitted externally. - Prompt injection artifacts: SKILL.md contains patterns that could try to override agent/system prompts. Manually review the raw skill docs for any instructions that attempt to change system-level behavior or tell the agent to ignore prior system instructions. - Secrets & tokens: Although no env vars are declared, several workflows reference GitHub API usage and other external integrations that will require tokens (GITHUB_TOKEN, service API keys). Do not supply repository or organization-level tokens to the skill unless you understand and accept the exact scope of operations. Prefer granting minimal-scope tokens and run initial scans in a throwaway/test repo. - File writes: The skill will read and write files under the repo (e.g., .rune/, assets/, AUDIT-REPORT.md). If you install in a sensitive repo, review writes first or run in a clone. Consider enabling version control hooks or running the skill in a sandboxed branch. - Least privilege: If you must use it, run the skill with least privilege (no org-scoped tokens, restrict network access if your environment supports it) and monitor outbound traffic. - Ask for provenance: If you want to proceed, request the author/homepage (the SKILL.md references github.com/rune-kit/rune and docs — verify those links resolve to the same code bundle). If you cannot verify the upstream repo and release, treat the package as untrusted. If you want, I can (1) list the specific files/lines that call external APIs or run shell commands, (2) search the code for any occurrences of 'process.env', 'GITHUB_TOKEN', 'curl', or exec, or (3) extract the raw SKILL.md and highlight the exact prompt-injection snippets for manual review.

Type: OpenClaw Skill Name: rune-kit Version: 2.15.0 The Rune skill bundle is a highly sophisticated and well-documented framework for AI-assisted development, implementing a mesh of 63 specialized skills. It features robust defensive mechanisms, including 'sentinel' for security gating, 'hallucination-guard' to prevent AI-generated import errors, and a dedicated 'integrity-check' skill (in skills/rune-integrity-check.md) designed to detect indirect prompt injection and memory poisoning in persisted state files. The architecture emphasizes process discipline through a 'skill-router' (in src/index.ts) and includes explicit risk disclosures for high-risk operations, such as the use of unofficial APIs in the Zalo extension. All high-privilege capabilities, such as file system access and command execution, are strictly aligned with the framework's stated purpose of providing a reliable, multi-session development environment.