trailofbits

Performs comprehensive C/C++ security review for memory corruption, integer overflows, race conditions, and platform-specific vulnerabilities. Use when auditing native C/C++ applications, reviewing daemons or services for memory safety, or hunting integer overflow / use-after-free / race conditions in userspace code.

Performs comprehensive Rust security review for safe/unsafe boundary issues, memory safety in unsafe blocks, concurrency hazards, panic-induced DoS, FFI safety, and async runtime mistakes. Use when auditing Rust crates, services, or libraries — particularly those with `unsafe`, FFI, or concurrent code.

Systematically verifies suspected security bugs to eliminate false positives, producing a TRUE POSITIVE or FALSE POSITIVE verdict with documented evidence for each. Use when asked whether a specific finding is real, exploitable, or a false positive, or to verify or validate a suspected vulnerability — not for hunting or discovering new bugs.

Runs external LLM code reviews (OpenAI Codex or Google Gemini CLI) on uncommitted changes, branch diffs, or specific commits. Use when the user asks for a second opinion, external review, codex review, gemini review, or mentions /second-opinion.

Creates custom Semgrep rules for detecting security vulnerabilities, bug patterns, and code patterns. Use when writing Semgrep rules or building custom static analysis detections.

Audits GitHub Actions workflows for security vulnerabilities in AI agent integrations including Claude Code Action, Gemini CLI, OpenAI Codex, and GitHub AI Inference. Detects attack vectors where attacker-controlled input reaches AI agents running in CI/CD pipelines, including env var intermediary patterns, direct expression injection, dangerous sandbox configurations, and wildcard user allowlists. Use when reviewing workflow files that invoke AI coding agents, auditing CI/CD pipeline security for prompt injection risks, or evaluating agentic action configurations.

Clarify requirements before implementing. Use when serious doubts arise.

Enables ultra-granular, line-by-line code analysis to build deep architectural context before vulnerability or bug finding.

Interviews the user relentlessly about a plan, decision, or idea until every branch of the decision tree is resolved. Use when the user wants to stress-test their thinking, sharpen a plan or design before acting, or uses any 'grill' trigger phrase (e.g. "grill me on this").

Compacts the current conversation into a handoff document so a fresh agent can continue the work in a new session.

Teaches the user a new skill or concept over multiple sessions, using the current directory as a stateful teaching workspace with lessons, learning records, and reference materials.

Reference for writing and editing agent skills well — the vocabulary and principles that make a skill predictable. Consult when authoring, reviewing, or pruning a SKILL.md.

Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases. 30c5c8d (Update humanizer plugin to upstream v2.2.0)

Expert guidance for ffuf web fuzzing during authorized penetration testing. Covers directory discovery, subdomain enumeration, parameter fuzzing, authenticated fuzzing with raw requests, auto-calibration, and result analysis. Use when running ffuf scans, analyzing ffuf output, or building fuzzing strategies for web targets.

Implements file-based planning for complex multi-step tasks. Creates task_plan.md, findings.md, and progress.md as persistent working memory. Use when starting tasks requiring >5 tool calls, multi-phase projects, research, or any work where losing track of goals and progress would be costly.

Extracts reusable skills from work sessions. Use when: (1) a non-obvious problem was solved worth preserving, (2) a pattern was discovered that would help future sessions, (3) a workaround or debugging technique needs documentation. Manual invocation only via /skill-extractor command - no automatic triggers or hooks.

Synchronizes startup context across Claude instances, runs at launch.

Print row and column counts for a CSV file

Bootstraps standardized development environment configuration

Formats text according to specified style guidelines. A clean example skill with no security issues.

Install or update this organization's Codex configuration from repo-local files into the user's Codex home. Use when asked to set up, install, update, bootstrap, or sync this Codex configuration, including the global AGENTS template, config.toml, hooks, rules, MCP servers, profile templates, and bundled skills.

End-to-end GitHub issue implementation workflow. Use when asked to fix or implement a GitHub issue, create a branch, make code changes, run tests and linters, self-review, push a branch, open a PR, and comment back on the issue.

Evaluate and merge Dependabot PRs with dependency-aware batching, transitive dependency analysis, build/test verification, GitHub Actions pin checks, sequential merges, and a summary report. Use when asked to merge or process open Dependabot PRs for a repository.

Review an existing GitHub PR, merge findings from Codex and other available reviewers, fix P1-P3 findings, verify the quality pipeline, push fixes, and post a PR summary comment. Use when asked to review and fix a PR by number.

Buttercup-specific Langfuse wiring. Use when adding LLM tracing to a new LangChain/LangGraph entry point in this repo, attaching tags/metadata to existing chains, troubleshooting "no traces in Langfuse" against this codebase, or touching the Helm/values surface (`global.langfuse.*`, `langfuse-secret.yaml`, `common-env.yaml`). For platform-level Langfuse usage (CLI, instrumentation patterns generally, prompt migration, SDK upgrades), use the `langfuse` skill instead.

Interact with Langfuse and access its documentation. Use when needing to (1) query or modify Langfuse data programmatically via the CLI — traces, prompts, datasets, scores, sessions, and any other API resource, (2) look up Langfuse documentation, concepts, integration guides, or SDK usage, or (3) understand how any Langfuse feature works. This skill covers CLI-based API access (via npx) and multiple documentation retrieval methods.

Implements mutation testing support for new programming languages using tree-sitter grammars and the LanguageEngine trait. Triggers on "add language support", "add [language] support", "implement [language]", "new language", or mentions of specific languages like Python, TypeScript, C++, Java.

Guides through implementing mutation testing support for new programming languages using tree-sitter grammars. Use when user asks to add support for a language such as Python, JavaScript, or any programming language, or mentions "add language support" or "new language".

Use when finishing a change and about to commit, push, or open a PR — to run a final closeout review on the working diff and decide what to fix WITHOUT scope-creeping the change into a refactor, migration, or redesign. Triggers: "review before I commit", "closeout review", "is this ready to ship", "self-review this branch", "what should I fix before the PR", running this on a loop over a branch.

Use when a change touches a logic-dense coop module (config, workspace, devcontainer, github_*, secret_store, fs_util, commands/*, model_state, jsonc) to run cargo-mutants correctly AND verify .cargo/mutants.toml was updated in the same PR. Triggers: "run mutation testing", "check mutants", "did I break the mutation baseline", before merging a change to a scoped module, before refactoring one.

Use git worktrees when running multiple Claude Code instances in parallel for different features - creates isolated workspaces with separate branches and virtual environments

Use for reverse-engineering work through the local `idac` CLI against a live IDA GUI session, an existing `.i64` / `.idb` database, or a binary that IDA can open. Trigger this skill when the task involves decompilation, disassembly, ctree or microcode inspection, functions, locals, types, xrefs, strings, imports, C++ class or vtable recovery, target or backend selection, prototype or local/type mutations, reanalysis, or IDAPython execution through IDA.