Each directory is one skill: a SKILL.md an agent loads on demand, with any
references/ it needs.
| Skill | What it does |
|---|---|
mantis-advise |
Proactive security advisor and architectural remediation assistant for secure code development. |
mantis-architecture |
Synthesizes raw learnings and codebase analysis into an interlinked Markdown Knowledge Base (KB). |
mantis-calibrate |
Calculates the final risk score based on empirical evidence and architectural impact. |
mantis-chain |
Analyzes individual security findings to identify and construct complex exploit chains. |
mantis-critic |
Assesses the production viability of findings, filtering out debug-only features and assertion traps. |
mantis-dedupe |
Consolidates raw security findings to eliminate redundant reports. |
mantis-history |
Analyzes the repository's version control system (VCS) history to extract past vulnerabilities, security fixes, and vulnerability patterns. |
mantis-meta-agent |
Acts as the persistent supervisor, launching and monitoring the automated review campaign. |
mantis-patch |
Generates minimal security fixes using transactional isolation (shadow directories or file backups), applies patches, and verifies them. |
mantis-pipeline-adapter |
Interactively guides the design and implementation of custom deterministic orchestrator harnesses. |
mantis-plan |
Formulates a targeted defensive security reviewing plan based on the active threat model and historical learnings. |
mantis-reflect |
Extracts learnings from execution trajectories at the end of a Mantis loop. |
mantis-report |
Generates a human-readable security review packet compiled from confirmed findings and exploit chains. |
mantis-reproduce |
Generates and runs crash reproducers to verify security flaws. |
mantis-researcher |
Audits production source code files based on the strategy in workspace/plan.json. |
mantis-review |
Independently reviews findings and filters out false positives. |
mantis-structural-index |
Builds a content-addressed semantic-unit index from source code for structural context. |
mantis-summarize |
Pre-processes the repository by generating security-focused summaries (mantis-summary.md) for each directory to make planning and research more efficient. |
mantis-threat-model |
Synthesizes trust boundaries, attack surfaces, and attacker profiles into a living threat model. |