Agent Skills by deduva
Showing 26 vetted skills indexed across 1 GitHub repositories.
{skill-name}
Load and validate SKILL.md frontmatter metadata for agent discovery.
agent-collaboration
Coordinate squad agents with worktree awareness and decision recording.
reviewer-protocol
Lock out original authors from revising rejected work artifacts.
error-recovery
Apply structured error recovery patterns for squad agents across API, tool, and build failures.
squad-conventions
Enforce shared coding conventions across the Squad codebase.
session-recovery
Query the session_store to locate and resume interrupted Copilot CLI sessions.
ralph-two-pass-scan
Scan GitHub issues with lightweight fields, then hydrate selected matches.
personal-squad
Merges user-level AI agents from the global config into sessions while preventing writes to project state.
versioning-policy
Enforce strict semantic versioning across Squad monorepo packages and CI gates.
economy-mode
Shift Layer 3 model auto-selection to cost-optimized alternatives when economy mode is active.
cross-machine-coordination
Coordinate distributed squad tasks across machines via git-based queues.
history-hygiene
Record final decision outcomes to history.md and update entries on reversals.
iterative-retrieval
Enforce three-cycle sub-agent spawns with WHY-context and coordinator validation.
notification-routing
Route notifications by type to dedicated channels via a dispatcher and adapter.
pr-lifecycle
Automate GitHub PR lifecycle from issue pickup to merge.
reflect
Capture and classify conversation patterns into structured learnings for future sessions.
init-mode
Propose a team roster and create a .squad directory after user confirmation.
test-discipline
Update tests to align with changed APIs and disk state in commits.
windows-compatibility
Enforce Windows-safe paths and CLI command patterns in scripts.
release-process
Automate pre-release validation and npm publish commands for multi-package repositories.
external-comms
Draft community responses for open issues and route them for human approval.
nap
Compress agent histories and prune old logs to reclaim context window budget.
humanizer
Enforce warm, human-sounding tone in PAO external communications.
cross-squad
Coordinate cross-squad work via manifest-based discovery and safe handoffs.