# Hunt Ssti

> Investigate authorized server-side template injection boundaries. Use when typed discovery or target-origin behavior identifies server-rendered inputs, evaluated-expression differentials, template-engine errors, rendering workflows, or suspected Jinja, Twig, FreeMarker, Mako, ERB, Smarty, Thymeleaf, or Django template evaluation.

- Skill: `duriantaco/hunt-ssti` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add duriantaco/hunt-ssti`
- Raw SKILL.md: https://api.skillmd.com/api/skills/duriantaco/hunt-ssti/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Product & Planning
- Author: duriantaco (https://skillmd.com/u/duriantaco)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/duriantaco/hunt-ssti

---


# Hunt Template Evaluation Boundaries

Treat this card as advisory prioritization. Reflection and template syntax do not prove server-side
evaluation.

## Workflow

1. Preserve a target-observed server-rendered request and its render transition.
2. Pair a bounded expression with a baseline and same-shape syntax control; change one input.
3. `ssti_fingerprint` may escalate to fixed file, environment, and command candidates. Run it only
   when the complete set is authorized; use `command_boundary` only on an independent typed sink.
4. Preserve identity, route, fields, controls, and evidence references. Report validated
   vulnerabilities independently of any challenge objective.

## Evidence Gate

Confirm only a derived expression result absent from the paired control and replayable at the same
sink. Reflection, syntax disappearance, engine errors, and framework names remain observations.
Keep any `contract_missing` result suspected rather than promoting it.

## Stop Conditions

Stop without a server-rendered replay, reproducible transition, clear control, or policy budget.
Do not add arbitrary code or extraction beyond the native set; redact environment data and bodies.