portal

Persona at invocation

Before conversational narration, read Persona: and Style: in the active project's ## Agent skills block from CLAUDE.md or AGENTS.md. Prefer the file containing the block, then an existing file; ties use CLAUDE.md. A symlink pair is one file. Read the saved value anew on each invocation, including from a subdirectory using the project root. No accessible project or no line means ordinary behavior. Do not search another project or global settings for this preference.

During the Persona at invocation stage, archmage on either line loads this skill's own references/archmage.md for the active workflow. off or an absent value leaves ordinary behavior active. An unknown value leaves ordinary behavior active and gets a brief explanation when conversational output is allowed; it does not stop the work. Explicit conversation instructions override the saved voice without writing settings. A request to enable Archmage for this workflow also loads the local reference.

Apply the voice only to lead-agent conversation. Deliverables, specialist roles, reply-only responses, and JSON-only output retain their contracts, with no added narration. End the persona with this workflow unless the user requests otherwise or a Style: line names archmage, which keeps the voice on for the whole session.

Portal

Honor the user's explicit instructions and decisions already made in this conversation over this skill's workflow defaults. Continue authorized work; ask only about unresolved choices that would materially change the result. Preparing or reviewing work does not authorize publishing it.

If a skill rule requires a pause or leaves requested work unfinished, name and link to the exact SKILL.md and quote the rule. Then explain what decision or prerequisite is missing. Distinguish a required gate from your interpretation.

Portal recommends one next action from the board or a named issue. Discovery is read-only: leave claims and other writes to the authorized handoff. A recommendation-only request ends with the report. Otherwise ask once before handoff, unless go or prior conversation already authorizes the selected action. With no actionable route, report the reason and stop without a question.

Optional run continues one explicitly named map or build effort serially through Stage Run. Equivalent explicit natural language selects the same path. It never claims the global queue, starts implementation after a map destination, or merges PRs. Build units use distinct owned worktrees and branches. Saved state grants no authority. Bare routing and go retain one handoff.

After an authorized handoff, continue the selected task to its requested completion, preserving the user's scope and prior decisions. A missing sibling skill does not end authorized work; use the ordinary-task fallback in Stage 5.

<SKILL_DIR> is the absolute directory this SKILL.md lives in. Substitute the real path every time it appears. Do not assign it to a shell variable first: a sandboxed or worktree-isolated session refuses bash "$VAR/script.sh" because it cannot resolve the path to read the script.

Arguments

Input	Effect
none	Read the whole board and the current branch, then route (Stage 2)
issue id or URL	Route that one issue (Stage 3)
`run <map-or-build-effort>`	Continue authorized units of this named scope through Stage Run
`go`	Skip the Stage 5 question and step through the route at once. Same meaning as the user saying "just do it" or "go"

An id is whatever the tracker uses (42, ENG-42, PLAT-42). A pull request number counts as an id when the tracker shares a number space with pull requests.

Execution spine

Resolve the tracker and the label strings (Stage 1).
Explicit named continuation: Stage Run. Otherwise blank argument: Stage 2. An id or URL: Stage 3.
Write the report (Stage 4).
Finish with the report or perform the authorized handoff (Stage 5).

Stage 1: Tracker

Read docs/agents/issue-tracker.md when it exists. Its Tracker: line names the tracker and its Adapter flags: line gives the flags for the bundled script. Missing file: GitHub, no flags. On a GitHub Enterprise host, pass GH_HOST=<host> inline on every script call. Shell state does not persist between calls.

Outside Stage Run, portal uses only the read side of scripts/tickets.sh: list, view, body, children, find, blocked, and next without --claim. Map reads use scripts/map.sh on GitHub (frontier, children, view, parent). On another tracker, the tracker file's "Wayfinding operations" section says what a map and a frontier are there; read them through that connector or API. For local, tickets are files: read them directly. For other, follow the tracker file's Conventions.

Resolve the label strings once. When docs/agents/triage-labels.md exists, take the right-hand column for ready-for-agent, needs-triage, needs-info, and ready-for-human; otherwise those are the strings. Map labels are scry:map and scry:<type>, with wayfinder:map and wayfinder:<type> as legacy spellings that mean the same thing.

Stage Run: Continue a named effort

Load references/run.md only for explicit named continuation. Resolve scope and current authority, lock and reconcile local state, then run its serial live-frontier loop. Its stop report replaces Stages 2 through 5. A single-ticket request keeps Stage 3 and its existing completion boundary.

Stage 2: Read the board

For a blank target, load references/board.md. Gather the board summary, then inspect route candidates in precedence order. Independent reads may run together. Resolve the chosen route before handing off.

Stage 3: Route one issue

For an id or URL, load references/issue.md. Inspect only that target and the relations needed to route it. Do not run whole-board discovery to fill the report.

During either routing stage, load references/build.md only when evaluating an effort's available ticket or progress. Load references/pr.md only when evaluating a pull request's feedback or checks.

Stage 4: Report

Write the result as markdown, not as a code block and not as plain indented lines. A reader scans this in a terminal that renders markdown, so the fields go in a table and the links stay clickable. Plain sentences, no dashes, under this repo's dispel rules when that skill is installed. Refer to every issue by its title with the link wrapped inside the name; a bare number appears only in the prompt line. Emit it exactly in this shape, including the empty header cells:


Branch	name, then clean \| uncommitted \| unpushed, then the PR state or `no PR`
Maps	count open and the first frontier ticket, or `none`
Build effort	title, done/total, then held by you \| available \| blocked \| awaiting review counts, or `none`
Roadmap	current milestone and done/total, or `none`
Ready	the oldest ready ticket nobody holds, or `none`
Inbox	count needing triage

Next step: <skill> on <ticket title with link>. <One sentence on why this row won.>

Prompt: \<one line that starts it>

For whole-board mode, every row appears in this order. Use not inspected for deliberately deferred detail, none for a verified absence, and unknown: <reason> for a failed read. For a named target, replace the board rows with Target, State, and Relevant context; include only the target and relations actually inspected. When no action is available, replace Next step with the reason and omit Prompt. A cell holds one line: no newlines, no bullets, and a literal pipe inside a value is escaped as \|. Links go in bare so the terminal renders them. A read the tracker refused is written as unknown: <reason> in its row, and the route is chosen from what did load.

The prompt line is natural language that works whether or not the named skill is installed. Shapes:

Implement <ticket URL>, following its brief. for a build ticket.
Resolve <ticket URL> on its map. for a map ticket.
Check progress and close out the build effort at <parent URL> once the pending work is complete. for a stalled effort.
Use the completed planning map <map URL> to propose the implementation work, keeping the map as the planning source. for a map with no effort.
Triage <issue URL>. for the inbox.
Report the roadmap and the next milestone to start. for a clear board with a roadmap.
Review the pull request <PR URL>. or Resolve the review feedback on <PR URL>. for a PR.

Stage 5: Step through

The report is delivered before this question, in a user-visible message, so the choice is about action. For an actionable route that still needs authorization, ask one question. Skip it for a recommendation-only request, go, or prior authorization covering this action. Use a question tool only when it is available and permitted in the current mode; otherwise ask in conversation and wait for the reply. Do not discover or call a tool by a name from another platform. Begin it with "Step through to <skill> on <ticket title>?" and offer these options, the first marked recommended:

Run <skill> on <ticket> now. Hand off as described below.
Take the runner-up instead. Offer this only when another actionable route was verified during discovery. Name it. Do not fetch more tickets just to populate this option. Choosing it hands off to that route the same way.
Something else. Stop with the report and the prompt line in view. The person types what they want.

A pending question is not authorization. Resume the chosen route after the answer without repeating the question.

Handing off. The routed skill is a sibling of this one: its instructions live at <SKILL_DIR>/../<skill>/SKILL.md. Read that file in full and follow it from its start as if the person had invoked it with the ticket id as the argument, including its persona and tracker stages and any applicable claim. Carry forward the selected branch or PR URL when the route is not an issue. For scan followed by reveal, preserve both steps when the user authorized that sequence. Portal has not claimed anything, so any claim belongs to the routed workflow. Do not summarize the sibling's rules from memory; the file is the contract. When the sibling folder is missing, say so, then carry out the prompt line as an ordinary task and note that the skill's own workflow, such as proof capture and the pull request, is not in play.

go with a route that leads to nothing (a clear board, a ticket held by someone else, or a ticket waiting on a person) prints the report and stops; there is nothing to step through.

References

Reference	Load at	Purpose
`references/run.md`	Stage Run, explicit named continuation	Scoped serial units, workspace isolation, reconciliation and stop report
`references/board.md`	Stage 2, blank target	Board summary and route precedence
`references/issue.md`	Stage 3, named target	State-first issue routing
`references/build.md`	Stage 2 or 3, effort candidate	Available ticket and progress
`references/pr.md`	Stage 2 or 3, PR candidate	Review and check evidence
`../<skill>/SKILL.md`	Stage 5, authorized handoff	The routed skill's own instructions, followed from the top