Issues

ctx.sinks is the kernel's own SinkRegistry, so any plugin can export to, read, close and replace a neighbour's configured sink

neutral:fix

#1961

· philcunliffe opened

on Sep 20, 2026

A plugin-lock entry missing install_dir crashes every kernel-booting hyp command with a bare TypeError naming no file and no entry

neutral:fix

#1958

· philcunliffe opened

on Sep 19, 2026

A stale lock entry whose install_dir was deleted is told 'its manifest will not load', which is imprecise when the directory itself is gone

#1957

· philcunliffe opened

on Sep 19, 2026

pluginRepairsByPointer clamps the install directory at 200 chars, which could truncate a pathological path out of the doctor repair line

#1956

· philcunliffe opened

on Sep 19, 2026

hyp config validate still says an installed plugin 'is not installed' when its manifest will not load, contradicting hyp status, plugin list and plugin info

neutral:fix

#1954

· philcunliffe opened

on Sep 19, 2026

ctx.sources.get(name) hands a plugin a neighbour's start() function, so a neighbour's source still runs under a squatter's context

neutral:fix

#1953

· philcunliffe opened

on Sep 19, 2026

PR #1949 deferred F1: selfUpdateStateRoot re-derives HYP_HOME instead of reusing the shared helper

#1952

· philcunliffe opened

on Sep 19, 2026

The owner-to-context selection rule is written three times and the gateway copy has no test

#1948

· philcunliffe opened

on Sep 19, 2026

CI runs no hermetic smokes, so a red flow can sit on master unnoticed (proposal: run the smoke tier on PRs)

#1941

· philcunliffe opened

on Sep 19, 2026

config_load_validate smoke is red on master: the grep compatibility entry inflates the reported plugin count to 3

#1940

· philcunliffe opened

on Sep 19, 2026

ctx.signal is never populated, so the abort guard in all three backfill providers is dead code

#1935

· philcunliffe opened

on Sep 19, 2026

parseCredentialHelperScript swaps roles for a hypothetical wrapper baking a non-absolute CLI

#1928

· philcunliffe opened

on Sep 19, 2026

Navigation Menu

All issues

ctx.sinks is the kernel's own SinkRegistry, so any plugin can export to, read, close and replace a neighbour's configured sink

A plugin-lock entry missing install_dir crashes every kernel-booting hyp command with a bare TypeError naming no file and no entry

A stale lock entry whose install_dir was deleted is told 'its manifest will not load', which is imprecise when the directory itself is gone

pluginRepairsByPointer clamps the install directory at 200 chars, which could truncate a pathological path out of the doctor repair line

hyp config validate still says an installed plugin 'is not installed' when its manifest will not load, contradicting hyp status, plugin list and plugin info

ctx.sources.get(name) hands a plugin a neighbour's start() function, so a neighbour's source still runs under a squatter's context

PR #1949 deferred F1: selfUpdateStateRoot re-derives HYP_HOME instead of reusing the shared helper

The owner-to-context selection rule is written three times and the gateway copy has no test

CI runs no hermetic smokes, so a red flow can sit on master unnoticed (proposal: run the smoke tier on PRs)

config_load_validate smoke is red on master: the grep compatibility entry inflates the reported plugin count to 3

ctx.signal is never populated, so the abort guard in all three backfill providers is dead code

parseCredentialHelperScript swaps roles for a hypothetical wrapper baking a non-absolute CLI

Search results