Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
/**
* GitHub Actions step-output/step-env/summary append helpers for CI scripts.
*
* ZERO external dependencies — these run in jobs that have not installed
* anything yet. Do NOT import from server/lib or any installed package.
*/
import { appendFileSync } from 'fs';
/**
* Append a `name=value` line to one of Actions' key/value command files.
*
* Newlines and carriage returns are stripped: the `name=value` form has no
* escape, so an embedded newline silently truncates the value and lets the
* remainder forge a second entry. Callers that render a value as markdown
* should sanitize further at their own call site.
*/
function appendCommandFile(envVar, name, value) {
const path = process.env[envVar];
if (!path) return;
appendFileSync(path, `${name}=${String(value).replace(/[\r\n]+/g, ' ')}\n`);
}
/**
* Append `name=value` to $GITHUB_OUTPUT, or do nothing outside Actions.
*
* @param {string} name - output name
* @param {unknown} value - stringified before writing
*/
export function writeStepOutput(name, value) {
appendCommandFile('GITHUB_OUTPUT', name, value);
}
/**
* Append `name=value` to $GITHUB_ENV, or do nothing outside Actions.
*
* Unlike a step output this needs no `id`, and later steps in the same job read
* it as an ordinary process environment variable — which is what lets one
* resolver step feed several plain `node scripts/...` steps.
*
* @param {string} name - environment variable name
* @param {unknown} value - stringified before writing
*/
export function writeStepEnv(name, value) {
appendCommandFile('GITHUB_ENV', name, value);
}
/**
* Append a markdown block to $GITHUB_STEP_SUMMARY, or do nothing outside
* Actions.
*
* Unlike the `name=value` files above, the summary is free-form markdown:
* newlines are content, not a delimiter, so they are preserved rather than
* collapsed. That makes the caller responsible for what it renders — pass
* fixed prose and values you control, never raw event payload text.
*
* @param {string} markdown - block to append; a trailing newline is added
* @param {NodeJS.ProcessEnv} [env] - injectable for tests
*/
export function writeStepSummary(markdown, env = process.env) {
const path = env.GITHUB_STEP_SUMMARY;
if (!path) return;
appendFileSync(path, `${markdown}\n`);
}
/**
* A workflow-sourced name (a job's, a step's) rendered into a log line.
*
* Actions parses a line beginning `::` as a workflow command, and the name
* comes from a workflow file, so a newline or a `::` would let a name forge a
* command instead of appearing in one. Capped because these land in an
* annotation title, not in a report.
*
* @param {unknown} value
* @param {string} [fallback] - used when the value is empty
* @returns {string}
*/
export function safeWorkflowText(value, fallback = '') {
return String(value || fallback)
.replace(/[\r\n]+/g, ' ')
.replace(/::/g, ':')
.slice(0, 80);
}
/**
* Escaping for the two halves of a workflow command, which differ: a property
* value is delimited by `:` and `,` as well, so those need encoding there and
* must NOT be encoded in the message (where they are ordinary punctuation).
* `%` goes first, or it would re-encode the escapes that follow it.
*/
const escapeCommandData = (value) => String(value)
.replace(/%/g, '%25')
.replace(/\r/g, '%0D')
.replace(/\n/g, '%0A');
const escapeCommandProperty = (value) => escapeCommandData(value)
.replace(/:/g, '%3A')
.replace(/,/g, '%2C');
/**
* An `::error::` workflow annotation line, ready to print to stdout.
*
* Annotations are the one diagnostic surface that survives the run being
* cancelled out from under the job that wrote it: Actions records them the
* moment the line is printed, and renders them at the top of the pull
* request's Checks tab. That is why fail-fast reporting uses one — a job's
* step conclusions are readable over the API, but only by somebody who
* already knows to look (issue 7574).
*
* @param {string} title - short heading; shown in the Checks tab
* @param {string} message - the detail line
* @returns {string}
*/
export function formatErrorAnnotation(title, message) {
return `::error title=${escapeCommandProperty(title)}::${escapeCommandData(message)}`;
}