[Evolution] Создание агента incident-responder #111
Open
opened 2026-05-09 11:45:52 +00:00 by NW
·
1 comment
Labels
Clear labels
agent::agent-architect
agent::capability-analyst
agent::code-skeptic
agent::evaluator
agent::history-miner
agent::lead-developer
agent::orchestrator
agent::sdet-engineer
agent::system-analyst
agent::the-fixer
budget::exhausted
budget::sufficient
budget::warning
cascade::depth-0
cascade::depth-1
cascade::depth-2
cascade::depth-exceeded
cascade::depth-n
evolution::model-change
evolution::new-agent
evolution::new-skill
evolution::new-workflow
evolution::prompt-opt
memory::checkpoint
memory::fresh
memory::recoverable
memory::stale
permission::evolve-system
permission::read-only
permission::violation
permission::write-code
permission::write-config
phase::awaiting-review
phase::drafting-spec
phase::executing
phase::gathering-evidence
phase::refining-prompt
phase::verifying
priority::critical
priority::high
priority::low
priority::medium
quality::blocked
quality::fail
quality::needs-fix
quality::pass
size::l
size::m
size::s
size::xl
size::xs
status::blocked
status::cancelled
status::done
status::in-progress
status::new
status::planned
status::review
status::testing
type::bug
type::chore
type::documentation
type::enhancement
type::feature
type::refactor
type::test
Owned by agent-architect
Owned by capability-analyst
Owned by code-skeptic
Owned by evaluator
Owned by history-miner
Owned by lead-developer
Owned by orchestrator
Owned by sdet-engineer
Owned by system-analyst
Owned by the-fixer
Token budget exhausted
Token budget sufficient
Token budget low
No subagent calls
1-level subagent calls
2-level subagent calls
Depth limit exceeded
Unlimited subagent calls
Model change evolution
New agent evolution
New skill evolution
New workflow evolution
Prompt optimization evolution
Checkpoint stored
Checkpoint fresh
Checkpoint recoverable
Checkpoint stale
Can evolve system
Read-only access
Security violation
Can write code
Can write config
Agent awaits review
Agent is drafting specification
Agent is executing task
Agent is gathering data
Agent is refining prompts
Agent is verifying results
Critical priority
High priority
Low priority
Medium priority
Blocked by quality
Quality check failed
Needs fixes
Quality check passed
Large (4-8 hours)
Medium (2-4 hours)
Small (1-2 hours)
Extra large (>8 hours)
Extra small (<1 hour)
Blocked
Cancelled
Completed
Work in progress
New issue, not started
Planned for sprint
Under review
In testing
Something is broken
Maintenance task
Documentation
Improvement
New feature
Code refactoring
Testing
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: UniqueSoft/APAW#111
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Контекст
В ходе реального инцидента IR-2026-05-09 (3 сервера, RCE, ELF backdoor, web-shells) была выявлена уязвимость агентской сети:
Отсутствует специализированный агент для серверного incident response и system hardening.
Все задачи по очистке выполнялись orchestrator-ом через bash напрямую, что не масштабируется и не дает повторяемый pipeline workflow.
Цель
Создать универсального агента incident-responder, способного работать с произвольным сервером по предоставленному SSH-доступу — отчистка от malware, hardening, восстановление, отчетность.
Почему существующие агенты не подошли
Scope agent incident-responder
Обязательно умеет
Не делает
Permissions
Model
ollama-cloud/kimi-k2.6:cloud (аналитическая, большой context window, системные задачи)
Routing
Keywords: incident, compromise, breach, hack, malware, shell, backdoor, forensic, cleanup, harden, server security, remove virus, integrity check
Эволюционные задачи
Phase 1: Research
Phase 2: Specification
forensics-checklist.md— процедуры поиска malwaremalware-signatures.md— сигнатуры PHP shells, ELF backdoorshardening-procedures.md— CSF, fail2ban, AIDE, .htaccessbackup-verification.md— проверка целостности архивовserver-recon.md— определение ОС, панели, сервисовPhase 3: Implementation
Phase 4: Testing
Phase 5: Documentation
Status: phase::researching
Next agent: @capability-analyst
Related incident: IR-2026-05-09
Budget: 50000 tokens
Evolution complete: incident-responder agent created
Component Added
.kilo/agents/incident-responder.md.kilo/skills/incident-response/Files Created (7 new)
.kilo/agents/incident-responder.md.kilo/skills/incident-response/SKILL.md.kilo/skills/incident-response/forensics-checklist.md.kilo/skills/incident-response/malware-signatures.md.kilo/skills/incident-response/hardening-procedures.md.kilo/skills/incident-response/backup-verification.md.kilo/skills/incident-response/server-recon.mdFiles Modified (7)
kilo-meta.jsonkilo.jsonc.kilo/agents/orchestrator.mdincident-responderin task permissions and Task Tool table.kilo/capability-index.yamlincident_response -> incident-responder.kilo/KILO_SPEC.md@IncidentResponderto Pipeline Agents tableAGENTS.mdSecurity and Incident Responsesection.kilo/EVOLUTION_LOG.mdCommit
Verification Checklist
"*": deny+ subagent deniedincident-responder: allowStatus: READY
Next: Orchestrator now routes incident keywords (
incident,compromise,breach,hack,malware,shell,backdoor,forensic,cleanup,harden) to@IncidentResponder