skills/gsv-process-identity/SKILL.md
Guide on what a GSV process is, how to orient around its identity, cwd, virtual filesystem paths, source mounts, and runtime events.
npx skillsauth add deathbyknowledge/gsv gsv-process-identityInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Treat yourself as a durable GSV process, not as a browser tab, host terminal, or stateless chat session.
A process has:
init:1000 or task:<uuid>init, task, review, cron, mcp, app, or a package profileGSV is Linux-shaped but not POSIX. Paths, commands, and syscalls are the stable interface.
Use the native shell on target: "gsv" for these checks:
proc self
pwd
ls
skills list
man
Use proc list when you need sibling process state. Use Read on /sys/devices or the device list in prompt context before choosing a non-gsv target.
Use these paths as an orientation map before deciding which tool or target to use. On the native gsv target, these are virtual GSV paths, not host-machine paths.
/home/<user>: the user's durable home tree.~/context.d/*.md: short standing context loaded into process prompts.~/skills.d/*: reusable process skills; use skills show <skill> before relying on one.~/knowledge/: durable reference notes and knowledge pages; not loaded automatically.Put durable conclusions, artifacts, and handoff notes in files, package source, repositories, or knowledge. Do not treat active conversation history as the artifact of record.
/src/packages: visible installed package source trees mounted for the process./src/packages/<package>: inspect or edit package source when package source rules allow it./usr/local/bin: read-only command shims installed by packages.There is no implicit per-agent workspace path. Use the process cwd, explicit user files, package source mounts, repo operations, or a user-provided target path. Package source writes may be staged per process until committed with package source commands. Check package state before assuming edits are installed or shared.
/proc: live process inspection./proc/self: the current process./proc/<pid>/status: process label, pid, parent pid, profile, state, uid, gid, and groups./proc/<pid>/identity: JSON identity and current process metadata./proc/<pid>/context.d: assignment-supplied process context files./proc/<pid>/conversations: visible conversation ids for that process./proc/<pid>/conversations/<conversationId>/status: JSON conversation status./proc/<pid>/conversations/<conversationId>/history: current live conversation history as JSONL./proc/<pid>/conversations/<conversationId>/segments: compacted segment ids./proc/<pid>/conversations/<conversationId>/segments/<segmentId>: archived segment messages as JSONL.Use /proc/self for self-inspection. Use /proc/<pid> for sibling processes only when you have a reason and permission. Runtime history is inspectable state, not durable project output.
/var/spool/cron: visible Kernel schedules./var/spool/cron/<scheduleId>: JSON schedule definition./var/log/gsv/scheduler: recent scheduler run history as JSONL.These are read-only runtime views. Use sched commands or scheduler syscalls to create, update, run, or remove schedules.
/sys/config: readable system configuration./sys/users/<uid>: readable user-scoped configuration./sys/devices: registered devices and target capability metadata./sys/capabilities: group capability grants./etc: system manuals and stable operator reference material./etc/passwd, /etc/group, /etc/shadow: auth table views; write access is root-only where allowed./dev/null, /dev/zero, /dev/random, /dev/urandom: device-like virtual endpoints.Use /sys/devices and target metadata before choosing a non-gsv target. Use /etc and man for local operator reference.
Messages beginning with [Process Event]: are GSV runtime events, not ordinary user messages. Treat them as authoritative state updates about IPC replies, IPC timeouts, watched signals, schedules, compaction, resets, approvals, or lifecycle changes.
Do not quote the prefix back unless it is directly relevant.
Use target: "gsv" for GSV control-plane work, virtual filesystem paths, package commands, process operations, repo operations, and native shell commands.
Use a device target only when the file, command, credential, private network, OS package, or hardware dependency lives on that connected machine.
Native shell commands run inside the Gateway worker sandbox. They do not run on the user's laptop.
development
Use active GSV web shell browser targets to inspect windows/apps, run browser JS, open files, and move files across targets.
documentation
Guide on how to coordinate durable GSV processes, including spawning, IPC, handoffs, scheduled work, conversation state, and compaction.
documentation
Guide on how context and skills work in GSV and how to add/edit them.
testing
Guide on how to operate and update a GSV deployment, including which layer changed, what to validate, and what to deploy or sync.