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 tzachbon/smart-ralph --skill ralph-specum-triagegit clone --depth 1 https://github.com/tzachbon/smart-ralphWrote 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/tzachbon/smart-ralph/ralph-specum-triage)<a href="https://agentmods.dev/skills/tzachbon/smart-ralph/ralph-specum-triage"><img src="https://agentmods.dev/badge/skills/tzachbon/smart-ralph/ralph-specum-triage/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/tzachbon/smart-ralph/ralph-specum-triage"><img src="https://agentmods.dev/badge/skills/tzachbon/smart-ralph/ralph-specum-triage.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Agent Snooping · line 29 Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.Fix: Remove all code or instructions that list or read other skills' files or directories. Skills should operate independently; cross-skill access is a privilege escalation.
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 | $0.00049 | $0.01454 |
| Opus 5 | $0.00024 | $0.00727 |
| Sonnet 5 | $0.00010 | $0.00291 |
| Haiku 4.5 | $0.00005 | $0.00145 |
Grade A, and why
ralph-specum-triage scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 64 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Ralph Specum Triage
You are a coordinator, not a triage analyst -- delegate decomposition work to a triage-analyst sub-agent.
Contract
- Epic data lives under
specs/_epics/<epic-name>/ - Track the active epic in
specs/.current-epic - Do not guess on ambiguous epic or spec names
- Triage produces a plan for multiple specs. It does not implement them
Action
- Parse exact
--quickand exact--interactivetokens before routing. Reject both together,-q, variants, and natural-language substitutes. - Check
specs/.current-epic. If an active epic exists, load its.epic-state.jsonasSTATEand normalize mode before any possible question. With exact--quick, resume the matching active epic without prompting; an explicit different epic must include its goal. In interactive mode, summarize status and offer resume, details, or a new epic. - Validate the resolved epic name against
^[a-z0-9]+(-[a-z0-9]+)*$before constructing any path. Reject invalid names. Ifspecs/_epics/<epic-name>/.epic-state.jsonalready exists, reuse and resume it without replacing state or progress. If the directory exists without a valid state, stop and require a different name or an explicit user-authorized reset. Only a missing directory may be initialized. - Resolve or create only the epic directory,
.progress.md, and.epic-state.json. Serialize the name and goal through a JSON encoder such asjq --arg. Use.epic-state.jsonasSTATEfor every phase gate command with phasetriage. Use.epic-state.jsonasSTATEfor every writer. After that, do not require.ralph-state.json. Create noresearch.md,epic.md, or generatedplan.mdbefore approval. - For a new epic, run
scripts/phase_gate.py mode STATEwith the exact parsed mode flag, then registerspecs/.current-epicbefore discovery. Exact quick mode asks no setup question; missing required name or goal is an input error. - Resolve the output destination from explicit input and configuration. In exact quick mode, use local Spec files as the deterministic default and ask nothing. Only interactive mode may ask one administrative destination question when the destination remains ambiguous; that answer never satisfies the interview gate.
- Ensure skill discovery pass 1 exists for the epic goal. Collect plugin, project
.agents/skills, project.claude/skills, and current harness catalog entries. Select explicitly named skills and record shadowed duplicates. - Load
skills/interview-framework-codex/SKILL.md, its required algorithm and domain-modeling references, and every selected domain contract in both interactive and quick mode. In interactive mode, follow the algorithm for critical decomposition boundaries, stable cross-spec contracts, dependency choices, and sequencing risks. Ask no storage, naming, branch, or discoverable question. - In interactive mode, require explicit
approve and delegate; in exact quick mode, recordbypassed_quick. In both modes, runphase_gate.py check-delegationwith phasetriageand the current loaded-manifest identity before every artifact-producing child. Apply the shared hard-transition invariant before every triage artifact writer. A failedcheck-delegationin either mode stops this invocation before phase transition, child dispatch, or target-artifact write; after a normal-mode failure, only the next explicit invocation creates a fresh manifest/interview identity. After an exact--quickdelegation failure, the next explicit invocation reruns discovery, records a freshphaseSkillLoadand interview identity, and does not reuse a terminalbypassed_quickinterview or its discovery revision. Only a matching in-progresscollectingorawaiting_confirmationinterview is resumable. Exact--quickretains its existing question-and-approval bypass and discovery, manifest, delegation, and writer checks. - Read-only exploration children may inspect seams, constraints, and existing boundaries, but they write no epic artifact.
- Keep the existing epic-state, read-only exploration, and multiwriter flow, plus each child packet, identity tuple, and receipt behavior, unchanged. Use separate unique gated writer dispatches as needed:
- a triage research writer for
research.md - a triage plan writer for
epic.mdand generated specplan.mdfiles - a fresh revision writer for every artifact-review revision
Pass the absolute gate helper path, epic state path, identity tuple, unique teammate dispatch identity, and verbatim
phaseSkillLoadmanifest to each writer. Each writer records its manifest loads and passescheck-agent-writewith its unique identity before its first filesystem artifact write.
- The coordinator does not assemble, format, or revise artifact content. Validate the gated writer outputs and persist
.epic-state.jsonwith each spec, its status, dependencies, and gate receipts. - Keep
specs/.current-epicset to the active epic name. - Show the next unblocked spec and route back to
$ralph-specum-startfor per-spec execution.
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.
- 7d ago Changed 5e91bd072058
- 11d ago First seen · 64 lines · 49 tokens per session scan A 6799c0c387a2
ralph-specum-triage is a skill published in the GitHub repository tzachbon/smart-ralph (536 stars, last pushed today), licensed MIT. It adds 49 tokens to every session and 1,454 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 2026-08-30.
Other skills, from other repositories
remember
Review reusable project knowledge and decide what belongs in project memory, notepad, or durable docs.
docx-comment-reply
Reply to comments (批注) in Word .docx/.doc files: extract comment context, draft replies, write threaded replies back, and validate OOXML.
explaining-machine-learning-models
Explain trained machine learning models through feature attribution, local explanations, and behavior summaries. Use as an explicit/manual helper once a model already exists, not for training ownership, leakage auditing, or general ML strategy selection.
moai-ref-ui-polish
UI polish and interface-completion reference: the small visual details — concentric border radius, optical alignment, shadow-vs-border, motion easing, typography smoothing, tabular numbers, icon stroke weight, hit areas — that separate polished interfaces from generic ones. Agent-extending skill that amplifies…
config-validator
Validate AIWG configuration files and project setup for correctness and completeness.
aiwg-regenerate-copilot
Regenerate copilot-instructions.md for GitHub Copilot with vendor-specific content only.