artale
/
fable-agent
Template
1
1
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
fable-agent/COMMANDS.md

10 KiB
Raw Blame History

Command Reference

This document mirrors src/index.ts command registration and is the canonical CLI reference.

Trust boundary: live receipt > committed receipt > memory summary. Memory is advisory only; 8099/deploy via git-proxy is the guarded deploy route, and 8098/deploy-webhook is legacy/stale for fable-agent deploy trust.

Last sync: 2026-06-14.

Quick examples

fable-agent run "Review API security risks"
fable-agent session start "ship launch plan"
fable-agent skills list --tag security
fable-agent pai status
fable-agent fable5 stack "refactor module" --enable-plinius
fable-agent fusion run "summarize RFC" --openrouter
fable-agent plinius godmode "improve explanation quality"

Top-level commands

fable-agent <command> [subcommand] [options]

Command Primary purpose
run Execute full agent loop
verify Emit agentic verification attestation
attest Alias for verify
session Session lifecycle
skills Skill registry ops
state Knowledge state / stats view
security Security scanners
cyber Read-only defensive cyber evidence collection
pai PAI/bridge operations
status Runtime surface summary
workflow Load workflow JSON
benchmark Run/view benchmark metrics
cost Cost/usage reporting
exa Exa search utilities
demo Demo mode
learned Summaries of learned material
familiar Note capture and health graph
factory Factory/VPS status and deployment guards
forgejo Forgejo intake and GitOps task receipts
pai-pi PI integration mode
daemon Long-running mode queue/monitor
fable5 Upgraded execution stack
fusion Multi-panel execution
generate Media generation
plinius Fable 5 diagnostics and upgrade tools

Command matrix

verify / attest

  • verify <target>: verify a target and emit an agentic attestation
    • --out <path>
    • --run <id>
    • --scan-all
  • attest <target>: alias for verify
    • --out <path>
    • --run <id>
    • --scan-all

run

  • run <task>
  • Options:
    • -l, --loop <n>: feedback loop iterations
    • -w, --workflow <id>: workflow definition ID
    • -s, --skill <id>: explicit skill override
    • -c, --compound <n>: compound run count

session

  • session start <task>
  • session resume <id>
  • session list
  • session inspect <id>

skills

  • skills list
    • -t, --tag <tag>
  • skills create <name>
    • -s, --steps <steps>
    • -t, --tags <tags>
  • skills inspect <id>
  • skills evolve <id>
  • skills sync
  • skills sharpen

state

  • state
    • -t, --tag <tag>
    • --stats

status

  • status
  • Displays command availability, env bootstrap, and runtime wiring summary.

security

  • security ast <file>: summarize JS/TS structure and risky capabilities
  • security scan <target>: hidden Unicode, reversed tags, fake wrappers, role spoofing, and instruction-smuggling markers
    • --include-fixtures: include intentional red-team fixtures/generators

cyber

  • cyber preflight
    • --host <host>
    • --port <n>
    • --mode <mode>
    • --output <file>
  • cyber reconcile
    • --preflight <file>
    • --capabilities <file>
    • --output <file>

pai

  • pai status
  • pai telos
    • -s, --section <name>
  • pai proxy <prompt>
    • -m, --model <id> (default fi-groq)
    • -t, --temperature <n>
    • -s, --system <text>
  • pai sync
  • pai dream
    • -c, --cycle <n>
    • --dry-run
  • pai cert
    • -t, --task <text>
    • -l, --loop <n>
    • --skip-proxy-check
  • pai bridge
  • pai memory
  • pai packs
  • pai report
  • pai run <goal>

workflow

  • workflow <file>

benchmark

  • benchmark run
    • -b, --benchmark <id>
  • benchmark agents <task>: run the same prompt against built-in or JSON-defined harnesses
    • --with <csv>
    • --timeout-ms <n>
    • --harnesses <path>
    • --out <path>
    • receipts include benchmark hygiene metadata: split, contamination risk, and local-only claim scope
  • benchmark duel <task>: record a two-implementer eval winner
    • --a <label>
    • --b <label>
    • --winner <a|b|tie>
    • --reason <text>
    • --reviewer <human|agent>
    • --out <path>
  • benchmark duel-tally: tally two-implementer eval receipts
    • --dir <path>
  • benchmark set-grade: deterministic expected/actual set grader
    • --expected <csv>
    • --actual <csv>
    • --threshold <n>
  • benchmark trend
    • -b, --benchmark <id>
  • benchmark list

cost

  • cost report
    • -h, --hours <n>

exa

  • exa search <query>
    • -n, --num-results <n>
    • -t, --type <type>: keyword|neural|auto
    • --include-domains <domains>
  • exa company <name>
  • exa sync

demo

  • demo
    • -q, --query <q>

learned

  • learned
    • -v, --verbose

factory

  • factory check
    • --run <id>
  • factory capabilities
    • --file <path>
    • --run <id>
  • factory reconcile
    • --output <path>
    • --run <id>
  • factory rsi
    • --host <host>
    • --port <n>
    • --out <path>
    • --canary-proof <path>
    • --run <id>
  • factory verify-cycle
    • --host <host>
    • --port <n>
    • --skill <name>
    • --out <path>
    • --allow-dirty
    • --canary-proof <path>
    • --allow-reconciled-stale-factory
    • --run <id>
  • factory verify <target>: verify a target and emit an agentic attestation
    • --out <path>
    • --run <id>
    • --scan-all
  • factory attest <target>: alias for factory verify; emits an agentic attestation
    • --out <path>
    • --run <id>
    • --scan-all
  • factory deploy <command>
    • --token <token>
    • --gate-receipt <path>
    • --plan-receipt <path>
    • --url <url>
    • --out <path>
    • --yes
    • --run <id>
  • factory gate <task>
    • --repo <path>
    • --host <host>
    • --port <n>
    • --capabilities <path>
    • --diagnostic-only
    • --plan-receipt <path>
    • --run <id>

forgejo

  • forgejo intake <ref>
    • --repo <repo>
    • --out <path>
    • --dry-run
    • --run <id>
  • forgejo feedbackpilot <json>
    • --repo <repo>
    • --out <path>
    • --dry-run
    • --create-issue

familiar

  • familiar capture <note>
    • -s, --source <src>
  • familiar process
  • familiar graph
  • familiar health
  • familiar briefing
  • familiar loop
    • -p, --priority <n>
    • -d, --dir <path>

pai-pi

  • pai-pi status
  • pai-pi skills
  • pai-pi run <goal>
    • -l, --loop <n>

daemon

  • daemon start
    • -d, --detach
  • daemon stop
  • daemon status
  • daemon queue <goal>
    • -p, --priority <n>
    • -t, --tags <tags>

fable5

  • fable5 status
  • fable5 route <task>
    • -d, --domain <domain>
  • fable5 models
  • fable5 model-access
    • --output <path>
    • --run <id>
  • fable5 flue
  • fable5 channels
    • --run <id>
  • fable5 memory-b-cell <file>
    • --cache <path>
  • fable5 receipt-health
    • --root <path>
    • --window-minutes <n>
  • fable5 heartbeat
    • --id <id>
    • --name <name>
    • --status <status>
    • --task <text>
    • --completed <n>
    • --cost <n>
    • --run <id>
    • --url <url>
    • --secret <secret>
    • --out <path>
  • fable5 verify <task>
    • --repo <path>
    • --run <id>
    • --trace-out <path>
  • fable5 goal <text>
    • -i, --iterations <n>
    • -s, --min-score <n>
  • fable5 goal-audit <objective>
    • --output <text>
    • --require <label:evidence>
    • --context <text>
    • --out <path>
  • fable5 spec <task>
    • --repo <path>
    • --out <file>
  • fable5 autowiki
    • --repo <path>
    • --out <path>
    • --max-files <n>
  • fable5 zte <task>
    • --repo <path>
    • --out <file>
  • fable5 worktree <action> [name]
  • fable5 state <project>
    • --add-fact <text>
    • --add-rule <text>
    • --add-failure <text>
    • --add-lesson <text>
    • --approved <text>
    • --rejected <text>
    • --performance <text>
    • --repo <path>
    • --import
  • fable5 workflow <pattern> <task>
    • -s, --subtasks <list>
    • -n, --max-iterations <n>
    • -p, --panel <slug>
    • --category <name>
    • --intensity <level>
  • fable5 compound <lesson>
    • --skill <name>
  • fable5 chain <task>
    • --stages <list>
  • fable5 meta <task>
    • --stages <list>
  • fable5 teams <task>
    • --workers <list>
  • fable5 stack <task>
    • -p, --project <name>
    • --enable-worktrees
    • --enable-vision
    • --enable-workflows
    • --enable-fusion
    • --enable-plinius
  • fable5 rpc
    • -p, --port <n>

fusion

  • fusion run <task>
    • -p, --panel <slug>
    • -j, --judge <model>
    • --openrouter
  • fusion panels

generate

  • generate image <prompt>
    • -m, --model <id> (default fal-ai/nano-banana-pro)
    • --aspect <ratio>
    • -r, --reference <url>
  • generate video <source>
    • -m, --model <id> (default alibaba/happy-horse/image-to-video)
    • -p, --prompt <text>
    • -d, --duration <n>
    • --720p

plinius

  • plinius godmode <task>
    • -p, --panel <slug>
    • -m, --mode <mode>
    • --proxy
    • -v, --verbose
  • plinius ultra <task>
    • -o, --output <text>
  • plinius safety <input>
  • plinius parseltongue <input>
    • -c, --category <category>
    • -i, --intensity <level>
  • plinius autotune <domain>
    • -s, --scores <list>
  • plinius stm <text>
    • -m, --modules <list>
    • --max-length <n>
  • plinius refusal <output>
    • -m, --model <id>
    • -t, --task <text>
  • plinius observatory
    • -p, --provider <name>
    • -m, --model <name>
    • -t, --tag <tag>
    • -l, --limit <n>
  • plinius liberation
    • --min-success <n>
  • plinius transparency <task>
    • -m, --model <id>
    • --save

Env and runtime references

  • Env bootstrap command: loadEnv().
  • Env precedence: existing process env > .env values.
  • Files loaded from project root:
    • .env
    • .env.<NODE_ENV>
    • .env.local
  • Proxy-sensitive commands: pai proxy, pai cert, plinius godmode --proxy.

Parity checklist

Before merging command-layer changes, verify:

  1. COMMANDS.md includes each .command(...) addition/removal from src/index.ts.
  2. New or changed flags are listed in this file.
  3. CONFIG.md includes matching bootstrap/env/proxy updates when startup behavior changes.
  4. A smoke command confirms the command parses in the shell:
    • node dist/index.js <command> --help (or equivalent).