You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
polish: rewrite all 7 ambiguous notifications for humans
1. Watchdog: "no_improvement" → "hasn't improved after 350 uses"
2. New proposal: "Tool: exec, 347x" → "Your agent used exec 347 times"
3. Sub-pattern: "Native Tool Sub-Pattern" → "Your agent ran exec in docker domain"
4. Expired: "No response in 7 days" → "No action taken — automatically removed"
5. Revalidation: "targets native tool" → "exec is built-in, doesn't need a skill"
6. Blocked: "BLOCKED: credential leak" → "Failed security validation"
7. Upgrade blocked: same as #6
Every notification answers: what happened, why, what to do.
`Use: /forge approve ${canonicalName} or /forge reject ${canonicalName}`
238
+
`📋 ${bold("New Skill Proposal")}\n\n`+
239
+
`${bold(canonicalName)}\n`+
240
+
`Your agent ran ${mono(key)} in the ${sp.domain} domain ${sp.entries.length} times across ${spSessions.size} session${spSessions.size!==1 ? "s" : ""} (${Math.round(spSuccessRate*100)}% success)\n`+
`Use: /forge approve ${skillName} or /forge reject ${skillName}`
374
+
`📋 ${bold("New Skill Proposal")}\n\n`+
375
+
`${bold(skillName)}\n`+
376
+
`Your agent used ${mono(key)}${entries.length} times across ${sessions.size} session${sessions.size!==1 ? "s" : ""} (${Math.round(successRate*100)}% success)\n`+
return`${bold(a.skill)} hasn't improved after ${a.activations} uses — still at ${Math.round(a.successRate*100)}% (was ${Math.round(a.baselineRate*100)}% at deploy)`;
406
+
}else{
407
+
return`${bold(a.skill)} is declining — down to ${Math.round(a.successRate*100)}% success over ${a.activations} uses`;
0 commit comments