triage-alert
Skill2 starsChanged 18 days agoMITView on GitHub
Triage a security alert, scanner finding, or bug bounty report into a decision. Use when given a CVE alert, Dependabot notification, SAST finding, or vulnerability report that needs severity, reachability, and an action.
In this repository4 other files it gives its agents.
- cve-researchSkill
- harden-serviceSkill
- review-diff-for-vulnsSkill
- threat-modelSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 20
Check reachability first: Search the lockfile for the vulnerable version, then trace whether the vulnerable function is imported and called.
- Distinctive: line 47
Never repeat a CVE's severity without checking reachability yourself.
- Distinctive: line 49
If the alert cannot be verified because the code is unavailable, say so and stop.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Triage alert
Turn an unactionable alert into a decision: real, reachable, and urgent, or noise with evidence.
When to use
- A dependency scanner, SAST tool, or monitoring system raised a finding.
- A bug bounty report or user vulnerability disclosure arrives.
- Someone forwards a CVE and asks "are we affected".
Procedure
- Restate the claim in one sentence, with the component and version range it applies to. Done when: the restatement names a package or file and a version.
- Check reachability first. Search the lockfile for the vulnerable version, then trace whether the vulnerable function is imported and called. A dependency that is present but never invoked is not exploitable. Done when: the report states the resolved version and the call path, or says "not imported".
- Establish preconditions. What must the attacker control, and what privilege does the code path require? Done when: preconditions list is explicit, including "authenticated user" if applicable.
- Assign severity using the exploitability you verified, not the feed's score alone. If you disagree with the feed, say so and why. Done when: severity, rationale, and any deviation from the published score are recorded.
- Decide and record: patch now, patch in the next window, accept with a compensating control, or dismiss. Each decision names an owner-visible next action and a review date. Done when: the decision is one of the four, with a date.
Output format
## Claim
## Reachability
| package | resolved version | vulnerable range | imported | call path | evidence |
## Preconditions
## Severity
published: X assessed: Y why:
## Decision
action: owner: review-by:Rules
- Never repeat a CVE's severity without checking reachability yourself.
- Quote versions exactly as resolved in the lockfile.
- If the alert cannot be verified because the code is unavailable, say so and stop.
- Fingerprint secrets, never echo values.
Anti-patterns
- Patching a version range with no evidence the vulnerable code is reachable.
- Dismissing an alert because "we are too small to be targeted".
- Reporting a scanner's output as your own analysis.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

