Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add azixxxxx/Hydra --skill hydra-reviewgit clone --depth 1 https://github.com/azixxxxx/HydraWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/azixxxxx/hydra/hydra-review)<a href="https://agentmods.dev/skills/azixxxxx/hydra/hydra-review"><img src="https://agentmods.dev/badge/skills/azixxxxx/hydra/hydra-review/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/azixxxxx/hydra/hydra-review"><img src="https://agentmods.dev/badge/skills/azixxxxx/hydra/hydra-review.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00036 | $0.00792 |
| Opus 5 | $0.00018 | $0.00396 |
| Sonnet 5 | $0.00007 | $0.00158 |
| Haiku 4.5 | $0.00004 | $0.00079 |
Grade A, and why
hydra-review scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hydra Review
Principle
Return only findings that survive local verification. A fluent model claim without reproducible repository evidence is not a finding.
Workflow
-
Call
hydra_preflightfor the repository. Stop if authentication, provider routing, CLI capabilities, or platform checks fail. -
Call
hydra_budgetwith{"action":"status"}. Select Sonnet high for normal review. Select Opus xhigh for critical risk, architecture, or adversarial pre-merge review. Do not select Fable for ordinary code review and do not weaken a required review model to conserve allowance. -
Build the bounded context capsule below. Call
hydra_startwithaccessMode: "analyze",strictModel: true,budgetIntent.priority: "quality-floor", the capsule plus one reviewer role-specific delta, and a bounded turn count. -
Require structured output with a
findingsarray. Each item must contain:path: repository-relative file path;line: smallest relevant 1-based line number ornull;severity:critical,high,medium, orlow;claim: falsifiable description of the defect;evidence: code path, state transition, command, or test that supports it;uncertainty: explicit caveat ornone;suggestedCheck: narrow local verification step.
-
Record the job ID, inspect
hydra_status, and retrievehydra_resultonly after a terminal state. -
For each candidate finding, open the cited code and its callers. Attempt to disprove the claim. Run the suggested focused test or another safe reproduction when practical.
-
Drop any finding that local evidence does not support. Correct its severity when the stated impact or reach is unsupported.
-
Report verified findings first, ordered by severity. Include the exact path and line, impact, evidence, and remaining uncertainty. If none survive, state that no verified findings were found and name the checks performed.
-
Record
taskType: "review", scope, accepted claims, rejected claims, false positives,testsBefore,testsAfter, patch verdict, and user verdict when known withhydra_record_outcome.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 50 lines · 36 tokens per session scan A 5bc3ace02c6c
hydra-review is a skill published in the GitHub repository azixxxxx/Hydra (0 stars, last pushed 2d ago), licensed MIT. It adds 36 tokens to every session and 792 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
remarc
Manage Remarc session comments and contextual feedback. Use when the user mentions Remarc sessions, comments, handoff, triage, review, addressing, resolving, status updates, or summaries.
looper-qa
Use when a Looper-managed GitHub repo needs scheduled pre-merge QA — a PR carries the looper:qa label, the spec stage reaches looper:spec-ready, or the Looper reviewer loop requests an independent second pass. Runs the full QA cycle (Looper state probe → PR checkout → ffs code review → language-specific test suite →…
codeck
Route explicit requests from a host coding agent to one or more locally configured AI executors through Codeck, attach Markdown or other project files, moderate cross-model consultation, expose disagreements, and synthesize traceable results. Use when the user explicitly names Codeck or asks to consult, compare, or…
engine
Internal data-only review core invoked directly by /lens:review and by a programmatic orchestrator (matali) calling it as a compute-only step, to produce a review-findings JSON object from the session diff. Not user-invocable; returns data, writes nothing, never prompts.
review
Review the current session's changes against their spec and plan, adversarially verify the findings, and render an interactive review document. Use when the user wants to review what was just built before it ships, runs /lens:review, or asks to "review my changes / review this before I commit". Reviews Claude's own…
render-review
Internal pure-render entrypoint — given a review-findings object (plus optional prior findings, a diff ref, and intent), reconcile in-memory, assemble the review-model, and render the interactive HTML via walkthrough:render. Writes ONLY the output HTML; no lens state, no recompute, no task list. Invoked by an…