Skip to content

flurdy/agent-skills / skills/peer-session

peer-session

Verified

Ask or instruct an existing live agent session through native Claude Code messaging or Orca, locally or remotely. Requires exact recipient/message approval and correlated evidence; no automatic retries, transport switching, spawning or queueing.

FileKindSize
SKILL.mdskill md21.1 KB
tests/README.mdmarkdown2.1 KB
tests/test_contract.pyscript6.2 KB

tests/README.md

View on GitHub
# Peer-session evidence

Run `make test-peer-session` from the repository root (Python 3.10+, standard library
only). These offline instruction contracts check permissions, transport selection,
native binding and ping consent, message bounds, evidence wording, the preserved Orca
path and suite wiring. They do not call `ListAgents`, send messages, start Orca or
execute a model, so they are not live-delivery evidence.

## Manual native pilot

Needs Claude Code with cross-session messaging (v2.1.224+; v2.1.236+ in every session
for idle notices) and at least two disposable local sessions in one project. Every send,
pings included, keeps the skill's per-send approval.

1. From a third session, run `/peer-session tell <project> <status>`. Confirm self is
   excluded, nothing is guessed, and the numbered ping batch is approved before it is sent.
   With unrelated sessions also listed, confirm the approval offers only the intended
   project/session subset and stop—not a broader `ping all sessions` option. If working
   directories are withheld, confirm the subset explicitly before approving pings.
2. Choose one candidate and approve the final message. Confirm exactly one `SendMessage`,
   with no automatic resend or Orca fallback.
3. Run `/peer-session ask <project> <question>`. Confirm the reply's first line carries
   the request ID and that an idle notice alone is reported as response unverified.
4. Set one recipient's `crossSessionInbound` to `refuse`. Confirm the outcome is reported
   as refused or unverified and is never retried.
5. From a session on another machine with Remote Control connected, run
   `/peer-session ask <project> <question>` against a session listed as a Remote Control
   row. Confirm the approval shows its kind and machine label, the caller is never bound
   to its own Remote Control row, and the correlated reply arrives (or is reported as
   response unverified).
6. Repeat one Orca `ask` from Pi to confirm that path is unchanged.

Record the date, Claude Code and Orca versions, observed results and gaps in the owning
task. Static tests and same-session self-review never count as observed delivery.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall