cogni-ai-ou/Cogni AI Agent

Runs Cogni AI agent via OpenCode

Trust Signals

Scorecard Score: not yet scored
Maintenance Recency: Activelast commit May 31, 2026
License: MIT

Pinned Snippet

workflow.ymlSHA-pinned

uses: cogni-ai-ou/cogni-ai-agent-action@d2042fc8b1ac1e9c2259c1c4e9ce0967a84c6e7b # v1.0.0

tags can be moved; commit SHAs can't. why a SHA?

Inputs

name	description	required	default
agent	Agent to use.	no	—
force-run	Whether to force the agent to run regardless of condition checks.	no	false
gh-extensions	Comma-separated list of GitHub CLI extensions to install.	no	github/gh-aw,github/gh-models
mentions	Comma-separated trigger mentions for OpenCode.	no	/co,/cogni,/review,/brainstorm
mentions_bots	Comma-separated bot mentions for OpenCode.	no	@opencode-agent,@OpenCode-Agent
model	Model to use for OpenCode.	no	opencode/deepseek-v4-flash-free
oidc_base_url	Base URL for OIDC provider.	no	—
opencode-api-key	API key for OpenCode.	yes	—
open-telemetry	Whether to enable OpenTelemetry spans for AI SDK calls.	no	false
permissions	Permissions configuration for OpenCode. Supports hierarchical format with 'default' and agent-specific overrides (e.g., 'cogni-ai-agent-auditor', 'cogni-ai-architect', 'cogni-ai-brain-ops', 'cogni-ai-code-reviewer', 'cogni-ai-context7-ops', 'cogni-ai-devops', 'cogni-ai-docs-editor', 'cogni-ai-elite', 'cogni-ai-fact-ops', 'cogni-ai-github-ops', 'cogni-ai-keeper', 'cogni-ai-manager', 'cogni-ai-plan-reviewer', 'cogni-ai-python-dev', 'cogni-ai-security-auditor', 'cogni-ai-tester', 'cogni-ai-weaver', etc.). Example: default: bash: ls: allow cogni-ai-devops: bash: ansible: allow terraform*: allow	no	# Keep in alphabetical order. default: apply_patch: allow bash: &default-bash '': deny awk: allow base64: allow cat: allow continue: allow cp: allow echo: allow ex: allow find: allow # GH is allowed but restricted for the default agent. gh api graphql: deny gh issue: deny gh pr: deny gh: allow git: allow glob: allow grep: allow hexdump: allow jq: allow links: allow ls: allow mawk: allow mkdir: allow mv: allow pipenv: allow pipx: allow poetry: allow pre-commit: allow printf: allow rm: allow sed: allow set: allow stat: allow time: allow timeout: allow touch: allow true: allow vim: allow context7_query-docs: deny context7_resolve-library-id: deny edit: allow glob: allow grep: allow list: allow question: deny read: allow # Skills are limited to a curated list from the upstream cogni-ai-agent-skills repository. # Note: Changes here should be synchronized with agent-specific skill allowlists. # Note: Keep in alphabetical order. skill: &default-skill agent-log-analysis: deny agents-md-writer: allow agentskills: allow ansible: allow brainstorm: allow brainstorm-agent-runs: allow brainstorm-github-pr: allow cat: allow chrome-devtools: deny code-review: allow code-tour: allow codeql: deny coding-standard-writer: allow context-aware-ops: allow critical-thinking: allow datadog-agent: deny datadog-api: deny datadog-mcp: deny datadog-monitors: deny datadog-pulumi: deny devcontainer: allow dictation: deny direnv: deny docker: allow dockerfile: allow docs-review: allow docs-writer: allow dot-github: allow editorconfig: allow fact-writer: allow gdpr-compliant: allow gh: allow gh-agent-task: allow gh-api: allow gh-aw: allow gh-aw-compile: allow gh-aw-debug: allow gh-aw-firewall: allow gh-aw-firewall-debug: allow gh-aw-new: allow gh-codespace: allow gh-issue: allow gh-models: allow gh-pr: allow gh-run: allow gh-skill: allow gh-search: allow git: allow git-expert: allow git-filter-branch: allow git-merge: allow git-rebase: allow gitattributes: allow github: allow github-actions: allow github-ah: allow github-issue: allow github-pr: allow github-pr-review: allow github-script: allow github-topics: allow github-aw: allow github-aw-debug: allow github-aw-memory: allow github-aw-syntax: allow github-mcp-server: allow github-pulumi: allow mermaid: allow mermaid-beta: allow mcp-cli: deny minizinc: allow molecule: deny opencode: allow out-yaml: allow pdf: allow pipenv: allow pipfile: allow pre-commit: allow python: allow report-writer: deny robust-commands: allow sbom: deny security-audit: allow shell: allow skill-writer: allow subagent-task: allow tdd: allow tester: allow vim-ex: allow yaml: allow yq: allow # Keep explicit denies for built-in subagents to ensure they are omitted from task delegation lists. task: '': allow explore: deny general: deny todowrite: allow cogni-ai-agent-auditor: apply_patch: deny bash: <<: default-bash gh run view: allow glob: allow read: allow edit: deny skill: <<: default-skill agent-log-analysis: allow task: allow cogni-ai-architect: bash: <<: default-bash # Explicitly deny devops tools so legacy flat lists don't leak them ansible: allow ansible-galaxy: allow ansible-lint: allow ansible-playbook: allow docker: allow pulumi: allow terraform: allow terragrunt: allow task: allow cogni-ai-brain-ops: apply_patch: deny bash: # Brain-ops is a context-gathering agent; gathering context via GH/git is allowed. gh api: allow gh pr checks: allow gh pr view: allow gh run view: allow git log: allow git show: allow glob: allow read: allow edit: deny skill: <<: default-skill '': deny task: allow cogni-ai-code-reviewer: apply_patch: deny bash: <<: default-bash gh api: allow gh pr checks: allow gh pr comment: allow gh pr review: allow gh pr view: allow glob: allow read: allow edit: deny skill: default-skill cogni-ai-context7-ops: apply_patch: deny bash: deny context7_query-docs: allow context7_resolve-library-id: allow glob: allow read: allow edit: deny skill: default-skill task: allow cogni-ai-devops: bash: <<: default-bash ansible: allow ansible-galaxy: allow ansible-playbook: allow docker: allow molecule: allow pulumi: allow terraform: allow terragrunt: allow glob: allow read: allow skill: <<: default-skill molecule: allow task: allow cogni-ai-docs-editor: apply_patch: allow bash: deny glob: allow read: allow edit: allow skill: default-skill task: allow cogni-ai-elite: task: allow cogni-ai-fact-ops: apply_patch: deny bash: deny glob: allow read: allow edit: deny skill: default-skill task: deny cogni-ai-github-ops: apply_patch: deny bash: <<: default-bash gh api graphql: allow # Can also modify issues/discussions gh issue close: allow gh issue comment: allow gh issue create: allow gh issue delete: deny gh issue edit: allow gh issue lock: allow gh issue pin: allow gh issue reopen: allow gh issue transfer: allow gh issue unlock: allow gh issue unpin: allow gh pr close: allow gh pr comment: allow gh pr create: allow gh pr edit: allow gh pr lock: allow gh pr merge: allow gh pr ready: allow gh pr reopen: allow gh pr revert: allow gh pr review: allow gh pr unlock: allow gh pr update-branch: allow glob: allow read: allow edit: deny skill: default-skill task: allow cogni-ai-keeper: task: deny cogni-ai-manager: apply_patch: deny bash: deny glob: allow read: allow edit: deny skill: <<: default-skill agentskills: deny ansible: deny coding-standard-writer: deny devcontainer: deny docker: deny dockerfile: deny gh-agent-task: deny gh-aw: deny gh-aw-compile: deny gh-aw-debug: deny gh-aw-firewall: deny gh-aw-firewall-debug: deny gh-aw-new: deny gh-codespace: deny gh-models: deny gh-run: deny gh-skill: deny git-expert: deny github-actions: deny github-ah: deny github-aw: deny github-aw-debug: deny github-aw-memory: deny github-aw-syntax: deny github-mcp-server: deny github-pulumi: deny github-script: deny mermaid: deny minizinc: deny pdf: deny pipenv: deny pipfile: deny pre-commit: deny robust-commands: deny shell: deny skill-writer: deny vim-ex: deny task: allow cogni-ai-plan-reviewer: apply_patch: deny bash: deny glob: allow read: allow edit: deny skill: default-skill cogni-ai-python-dev: bash: <<: default-bash jinja2: allow python: allow python3: allow glob: allow read: allow skill: default-skill task: allow cogni-ai-security-auditor: apply_patch: deny bash: deny glob: allow read: allow edit: deny skill: default-skill task: allow cogni-ai-tester: bash: <<: default-bash molecule: allow npm test: allow pytest: allow glob: allow read: allow skill: *default-skill task: allow cogni-ai-weaver: task: allow
prompt	Prompt to pass to the agent.	no	""
variant	Model variant for provider-specific reasoning effort (e.g., high, max, minimal).	no	—
version_agents	Version of cogni-ai-agents to use.	no	main
version_instructions	Version of cogni-ai-agent-instructions to use.	no	main
version_skills	Version of cogni-ai-agent-skills to use.	no	main

Outputs

name	description
prompt	The resolved prompt sent to the agent.
response	The response from the agent.
run	Whether the agent was executed.