vouchington-tooling 0.23.0 → 0.24.0

package/skills/blackboard/SKILL.md

@@ -13,18 +13,38 @@ Use when the current repository provides a journal, blackboard, or equivalent du
 record. Read local `AGENTS.md` for the provider, credential, retention, and
 subagent-identity rules.
 
-1. Append a concise note when a check fails, permission is denied, scope changes, a repeated fix is
-   needed, or a tool result reveals a reusable gap. Include the finding, affected paths, concrete
-   evidence, and any tracking reference.
-2. Preserve the provider's required session and parent-session identity.
-   - Do not invent, guess, print, search for, or persist a credential outside its documented
-     mechanism.
-   - When the provider supports repository metadata, keep the cumulative canonical repository list
-     on that agent's own session. Tag each entry with only the repositories relevant to it.
-   - Update session metadata before appending. Stop if that update fails.
-3. If the journal service cannot authenticate or persist a mandatory entry, stop and report the
-   blocker rather than silently substituting a local file or memory.
-4. Read journal entries oldest-first when preparing a retrospective or clustering follow-ups; use
-   direct evidence rather than reconstructing events from memory.
+1. Capture consequential observations before filing an issue: failed or recovered checks, denied or
+   approved permissions, repeated fixes, scope changes, first-party tool behavior, and architectural
+   findings. Record useful resolved outcomes as well as unresolved gaps. Include bounded evidence,
+   affected areas, an explicit disposition, and an existing tracking reference when present.
+   Journal meaningful skips or bypasses of applicable configured workflows before completion, with
+   the owning workflow or command-catalog entrypoint, reason, and observed result/action.
+2. Use the shared validated envelope with storage `type: journal` or `retrospective`, canonical exact
+   entry repositories, stable `sourceEventId`, work outcome, and explicit feedback coverage. Category
+   is optional context; it never changes storage type or distillation eligibility. Generate routine
+   facts and metadata through the approved composer instead of hand-authoring factual markers.
+3. Preserve exact provider session and parent identities. Each agent writes only its own session.
+   Update that session's cumulative repository union before appending; never infer another repository
+   from prose or create a child identity from a model guess.
+4. Select the repository's trusted mode explicitly. Interactive outages may retain validated,
+   sanitized feedback in the bounded durable outbox. Report pending count and delivery state, replay
+   with the same source identity, and continue primary work only after persistence succeeds. A full,
+   unsafe, or unwritable outbox blocks capture; never evict or silently discard an unsent record.
+5. Autonomous runners require online session ensure, a fresh admission write, and readback before
+   launching an attempt. Terminal reporting preserves work outcome separately from feedback
+   coverage and delivery. A filesystem outbox never grants admission or acknowledged completion.
+6. Use at-least-once delivery with consumer deduplication by exact session/source identity. A
+   transport timeout can leave a late durable write; pending or blocked is never an acknowledged
+   receipt. Replay verifies matching content before retrying and rejects conflicting source reuse.
+7. Read journal entries oldest-first for retrospectives. Record explicit none-observed with inspected
+   scope, or not-assessed/unavailable with reason, when evidence does not establish a finding. A
+   permission request is not proof of approval or denial; localhost refusal is not proof of sandbox
+   enforcement. Mark partial capture and dropped counts instead of reporting clean evidence.
 
-Consumer wrapper owns: service, command, note grammar, archival policy, and issue destination.
+Never search for, print, invent, or mint credentials. Durable entries and outbox records contain only
+bounded structured facts or sanitized summaries, never raw command output, transcript content,
+environment dumps, provider payloads, or secrets. Known-secret redaction supplements the author's
+minimization responsibility; it cannot prove arbitrary prose contains no undisclosed secret.
+
+Consumer wrappers own service configuration, trusted mode selection, commands, outbox location,
+issue routing, and archival authorization.

package/skills/retrospective/SKILL.md

@@ -12,20 +12,31 @@ Local policy overrides these defaults; do not reload an already-read adapter or
 Use at task completion or session close-out when the repository has a retrospective workflow. Read
 local `AGENTS.md`, and journal guidance first.
 
-1. Check whether a retrospective already exists. Append only a material delta when local policy
-   says to append.
-2. Gather verifiable facts from the repository's approved status, validation, issue, and journal
-   sources. Do not estimate unknown facts from memory or transcripts.
-3. Summarize the outcome, plan-versus-actual differences, validation evidence, recurring friction,
-   and actionable process improvements. Keep one-off noise separate from a repeatable root cause.
-   Preserve each source entry's repository attribution when the journal supports it. A retrospective
-   that spans repositories names every represented repository.
-4. Save through the repository's required durable mechanism. Report the record identifier and any
-   follow-up decisions.
+1. Read the agent's own journal oldest-first and check for an existing retrospective. Capture a
+   material new observation before filing a follow-up issue. Preserve entry-level repository tags.
+2. Generate routine repository and transcript facts with the approved composer. Keep exact factual
+   section markers generated, preserve unavailable counts, and never replace missing evidence with
+   invented zero statistics. Do not paste raw collectors' output or manually reconstruct facts.
+3. Record work outcome, feedback coverage, and delivery independently. Success, failure,
+   cancellation, timeout, no-change, and policy refusal each need truthful terminal reporting.
+   Missing or truncated sources are unavailable/partial; durable interactive pending remains pending.
+   Autonomous completion requires the controller's acknowledged terminal readback or visible block.
+4. Report observed first-party tool and architectural behavior with concrete evidence and a
+   disposition, including useful resolved outcomes and rejected proposals. None observed requires
+   the inspected scope. Not assessed or unavailable requires a reason. Do not turn a plausible
+   cause into an observed diagnosis or infer permission decisions from requests.
+   For applicable configured tools, report used versus skipped/unavailable with reasons, the owning
+   workflow or command-catalog entrypoint, and terminal result/action, especially pr-shepherd and no-mistakes.
+5. Keep the top five findings or improvements in the human prose. The bounded durable evidence can
+   retain additional relevant findings; a prose limit is not permission to silently discard them.
+   Separate recurring root causes from one-off noise and preserve existing issue references.
+6. Save through the shared validated writer. Interactive outages use the approved bounded private
+   outbox with visible pending count; failed persistence blocks capture. Autonomous reporting stays
+   online and never uses that outbox to authorize an attempt or completion. Report the acknowledged
+   record/receipt, or the actual pending/blocked state and next replay action.
 
-Use raw evidence only for local verification. Save only bounded structured facts or redacted
-summaries in a durable retrospective; never embed unredacted logs, command output, environment
-dumps, provider payloads, or transcript content there.
-
-Consumer wrappers add transcript access, retention, a report schema, a journal provider, and issue
-filing policy. They cannot weaken this minimization boundary.
+Use raw evidence only for local verification. Save only bounded structured facts or redacted summaries
+in durable retrospectives; never embed unredacted logs, command output, environment dumps, provider
+payloads, or transcript content there. Exclude secrets. Consumer wrappers own providers, commands,
+report schema, retention, trusted runner mode, and issue routing; they cannot weaken this minimization
+boundary.

package/skills/retrospective-distill/SKILL.md

@@ -15,17 +15,26 @@ Use when completed retrospectives or journals should become durable follow-up wo
 1. Enumerate only completed, eligible session records. For repository-scoped work, filter sessions
    by repository membership and use only entries attributed to that repository. Leave untagged
    legacy records unclassified. Do not infer their repository. Leave in-progress sessions intact.
-2. Cluster findings by root cause. Prefer a few broad actionable themes over many narrow issues.
+2. Validate storage type and the shared envelope independently of optional category. Never repair
+   unknown legacy types or repository provenance by inference. Quarantine malformed sessions with
+   an explicit reason while processing valid unrelated sessions. Deduplicate at-least-once records
+   by exact session/source identity and reject conflicting duplicate content. Retain coverage,
+   dropped counts, requested versus observed outcomes, and useful resolved findings.
+3. Cluster findings by root cause. Prefer a few broad actionable themes over many narrow issues.
    Treat a finding already linked to an open tracker as context, not a duplicate.
-3. Verify each candidate against the current base. Search existing issues and open changes before
-   drafting. Skip work that is complete, explicitly rejected, or already covered.
-4. Draft self-contained issues with the problem, concrete proposed work, relevant areas, and
+4. Verify each candidate against the current base. Search existing issues and open changes before
+   drafting. Skip work that is complete, explicitly rejected, or already covered. A first-party version
+   bump is not proof of resolution: verify the adopted change against the original finding before
+   deferring. Capture a new observation in the journal before filing it.
+5. Draft self-contained issues with the problem, concrete proposed work, relevant areas, and
    validation. Route every authorized creation through
    [github-issue](../github-issue/SKILL.md), including its repository gate, label approval, and
    denied-external tracking behavior.
-5. Archive only records fully processed across every represented repository, under the repository's
+6. Archive only records fully processed across every represented repository, under the repository's
    retention rules. A one-repository pass leaves a multi-repository session active until the other
-   repositories are reviewed. Report created, updated, skipped, and deferred themes with reasons.
+   repositories are reviewed. Report reviewed, fixed, duplicate, deferred, quarantined, and
+   actionable dispositions with reasons. Keep the top five themes in human prose without dropping
+   underlying findings or unresolved records.
 
 Use source records only for local verification and leave them in the repository's approved journal
 or retention system. Public issue bodies contain only the minimum bounded facts or redacted